diff --git "a/attnserver.run_attnserver.slurm.sh.343221.out.log" "b/attnserver.run_attnserver.slurm.sh.343221.out.log" new file mode 100644--- /dev/null +++ "b/attnserver.run_attnserver.slurm.sh.343221.out.log" @@ -0,0 +1,8124 @@ +Running ctx_length=1024, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 1024 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 1024 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 1024 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 1024 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 1024 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None +INFO:megatron.training.initialize:Setting logging level to 0 + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True +INFO:megatron.training.initialize:Setting logging level to 0 + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. +INFO:megatron.training.initialize:Setting logging level to 0 + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.055 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.395 seconds +time to initialize megatron (seconds): 7.871 +[after megatron is initialized] datetime: 2025-06-21 21:26:34 +building GPT model ... +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 144247808 +>>> embedding +>>> embedding>>> decoder + +>>> output_layer +>>> decoder +>>> output_layer +>>> embedding>>> embedding + +>>> decoder +>>> decoder>>> output_layer + +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 144247808 > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 144247808 + + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 144247808 > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 144247808 + + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 144247808 +>>> embedding +>>> decoder +>>> output_layer +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 144247808 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 144247808 + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 144247808 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 144247808 +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (144247808 elements, 144247808 padded size): + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.final_layernorm.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.layers.0.mlp.linear_fc1.bias + module.decoder.layers.0.self_attention.linear_proj.weight + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.layers.0.self_attention.linear_proj.bias + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.embedding.position_embeddings.weight + module.embedding.word_embeddings.weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.bias +>>> embedding +>>> decoder +>>> output_layer +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 144247808 +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine +(TP, PP, encoder TP, encoder PP) mismatch after resume ((4, 1, 0, 0) vs (8, 1, 0, 0) from checkpoint): RNG state will be ignored +(TP, PP, encoder TP, encoder PP) mismatch after resume ((4, 1, 0, 0) vs (8, 1, 0, 0) from checkpoint): Rerun state will be ignored + loading distributed checkpoint from gpt-checkpoint at iteration 10 +Running ctx_length=2048, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 2048 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 2048 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 2048 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 2048 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 2048 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.044 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.466 seconds +time to initialize megatron (seconds): 8.981 +[after megatron is initialized] datetime: 2025-06-21 21:27:12 +building GPT model ... +>>> embedding>>> embedding + +>>> decoder>>> decoder + +>>> output_layer>>> output_layer + +>>> embedding +>>> embedding>>> decoder + +>>> output_layer +>>> decoder +>>> output_layer +>>> embedding>>> embedding > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 148442112 + + +>>> decoder +>>> decoder>>> output_layer + +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 148442112 + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 148442112 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 148442112 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (148442112 elements, 148442112 padded size): + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.final_layernorm.weight + module.decoder.layers.0.mlp.linear_fc1.bias + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.embedding.position_embeddings.weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.bias + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_proj.weight + module.embedding.word_embeddings.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_proj.bias +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine +WARNING: could not find the metadata file gpt-checkpoint/latest_checkpointed_iteration.txt + will not load any checkpoints and will start from random +(min, max) time across ranks (ms): + load-checkpoint ................................: (2.67, 3.60) +[after model, optimizer, and learning rate scheduler are built] datetime: 2025-06-21 21:27:13 +> building train, validation, and test datasets ... + > datasets target sizes (minimum size): + train: 10 + validation: 1 + test: 1 +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let mock = True, as both blend and blend_per_split are None +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split = 1,1,1, an arbitrarily even split, as mock is True +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 0.3333333333333333), (0.3333333333333333, 0.6666666666666666), (0.6666666666666666, 1.0)] +> building train, validation, and test datasets for GPT ... +INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building MockGPTDataset splits with sizes=(10, 1, 1) and config=GPTDatasetConfig(random_seed=1234, sequence_length=2048, blend=None, blend_per_split=None, split='1,1,1', split_matrix=[(0, 0.3333333333333333), (0.3333333333333333, 0.6666666666666666), (0.6666666666666666, 1.0)], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=True, tokenizer=, mid_level_dataset_surplus=0.005, reset_position_ids=False, reset_attention_mask=False, eod_mask_loss=False, create_attention_mask=True, drop_last_partial_validation_sequence=True, add_extra_token_to_sequence=True, object_storage_cache_path=None) +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset train indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.005938 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 33296 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset valid indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.002676 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 33281 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset test indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.002617 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 33343 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +> finished creating GPT datasets ... +[after dataloaders are built] datetime: 2025-06-21 21:27:13 +done with setup ... +(min, max) time across ranks (ms): + model-and-optimizer-setup ......................: (503.79, 528.37) + train/valid/test-data-iterators-setup ..........: (20.84, 139.52) +training ... +Setting rerun_state_machine.current_iteration to 0... +[before the start of training step] datetime: 2025-06-21 21:27:13 +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 0 +Done exporting trace 0 +Number of parameters in transformer block in billions: 0.35 +Number of parameters in embedding layers in billions: 0.21 +Total number of parameters in billions: 0.56 +Number of parameters in most loaded shard in billions: 0.1400 +Theoretical memory footprints: weight and optimizer=2403.18 MB + [2025-06-21 21:27:30] iteration 1/ 10 | consumed samples: 1 | elapsed time per iteration (ms): 16912.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 4294967296.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +[Rank 1] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3550.0 | max reserved: 3550.0 +[Rank 0] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3550.0 | max reserved: 3550.0 +[Rank 14] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3598.0 | max reserved: 3598.0[Rank 10] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3438.0 | max reserved: 3438.0[Rank 8] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3406.0 | max reserved: 3406.0[Rank 13] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3534.0 | max reserved: 3534.0 + + + +[Rank 11] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3598.0 | max reserved: 3598.0 +[Rank 2] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3550.0 | max reserved: 3550.0 +[Rank 9] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3598.0 | max reserved: 3598.0[Rank 12] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3598.0 | max reserved: 3598.0 + +[Rank 4] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3566.0 | max reserved: 3566.0 +[Rank 3] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3550.0 | max reserved: 3550.0[Rank 5] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3630.0 | max reserved: 3630.0 + +[Rank 15] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3598.0 | max reserved: 3598.0 +[Rank 7] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3630.0 | max reserved: 3630.0 +[Rank 6] (after 1 iterations) memory (MB) | allocated: 2169.75439453125 | max allocated: 3251.16845703125 | reserved: 3630.0 | max reserved: 3630.0 +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192])batch tensor: +batch tensor: tokenslabels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +torch.Size([4, 8192])batch tensor: +attention_mask batch tensor:torch.Size([4, 1, 8192, 8192]) +labelsbatch tensor: torch.Size([4, 8192])position_ids + batch tensor:torch.Size([4, 8192]) +loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens batch tensor after cp:torch.Size([4, 2048]) +tokens batch tensor after cp: torch.Size([4, 2048])labels + batch tensor after cp:torch.Size([4, 2048]) +labelsbatch tensor after cp: torch.Size([4, 2048])loss_mask +batch tensor after cp: position_ids torch.Size([4, 2048]) + batch tensor after cp:torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor:loss_maskbatch tensor after cp: attention_masktorch.Size([4, 2048]) + tokenstorch.Size([4, 1, 2048, 8192])batch tensor after cp: + attention_maskbatch tensor after cp: position_idstorch.Size([4, 1, 2048, 8192]) +torch.Size([4, 2048])batch tensor after cp: +torch.Size([4, 8192]) position_ids +torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_maskbatch tensor after cp: torch.Size([4, 8192])tokens + batch tensor:torch.Size([4, 2048]) +attention_mask batch tensor after cp:torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +labels batch tensor:torch.Size([4, 2048]) position_ids + batch tensor after cp:torch.Size([4, 8192]) +loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 1 +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +Done exporting trace 1 +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + [2025-06-21 21:27:30] iteration 2/ 10 | consumed samples: 2 | elapsed time per iteration (ms): 102.6 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 2147483648.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels batch tensor:torch.Size([4, 2048]) +batch tensor after cp: loss_masktokens torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192])torch.Size([4, 8192]) + +batch tensor after cp: position_ids torch.Size([4, 2048])batch tensor: + labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 2 +Done exporting trace 2 + [2025-06-21 21:27:30] iteration 3/ 10 | consumed samples: 3 | elapsed time per iteration (ms): 75.8 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 1073741824.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids batch tensor:torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +Start exporting trace 3 +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +Done exporting trace 3 +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + [2025-06-21 21:27:30] iteration 4/ 10 | consumed samples: 4 | elapsed time per iteration (ms): 74.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 536870912.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor:batch tensor after cp: position_ids torch.Size([4, 2048])tokens + torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels batch tensor after cp:torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) + tokensbatch tensor: torch.Size([4, 2048])loss_mask +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) + batch tensor after cp:torch.Size([4, 8192]) +labels torch.Size([4, 2048])batch tensor: +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) + batch tensor after cp:attention_mask loss_masktorch.Size([4, 1, 8192, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor:batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192])position_ids +batch tensor: tokens torch.Size([4, 8192]) + batch tensor after cp:torch.Size([4, 8192]) +position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +Start exporting trace 4 +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +Done exporting trace 4 +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + [2025-06-21 21:27:30] iteration 5/ 10 | consumed samples: 5 | elapsed time per iteration (ms): 71.5 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 268435456.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor:batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +tokens batch tensor: position_ids torch.Size([4, 8192]) +torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp:batch tensor after cp: attention_mask tokens torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +torch.Size([4, 2048]) +batch tensor after cp: batch tensor after cp:position_ids labelstorch.Size([4, 2048]) +torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp:batch tensor: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: position_idsbatch tensor after cp: tokens torch.Size([4, 8192])torch.Size([4, 2048]) + +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +Start exporting trace 5 +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +Done exporting trace 5 +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + [2025-06-21 21:27:30] iteration 6/ 10 | consumed samples: 6 | elapsed time per iteration (ms): 70.8 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 134217728.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids batch tensor after cp:torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) + tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens batch tensor:batch tensor after cp: tokenstorch.Size([4, 8192]) +torch.Size([4, 2048])tokens + batch tensor:batch tensor after cp: labelstorch.Size([4, 8192])labels +torch.Size([4, 8192])torch.Size([4, 2048]) +batch tensor: +batch tensor: batch tensor after cp: labels loss_mask loss_mask torch.Size([4, 8192]) torch.Size([4, 8192]) +torch.Size([4, 2048]) +batch tensor: +batch tensor: batch tensor after cp: loss_mask attention_mask attention_mask torch.Size([4, 8192])torch.Size([4, 1, 8192, 8192])torch.Size([4, 1, 2048, 8192]) + + +batch tensor:batch tensor: batch tensor after cp: attention_mask position_ids position_ids torch.Size([4, 1, 8192, 8192])torch.Size([4, 8192])torch.Size([4, 2048]) + + +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: batch tensor:tokens tokens torch.Size([4, 8192]) +batch tensor:torch.Size([4, 8192]) labels +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) + batch tensor:torch.Size([4, 8192]) +labels batch tensor: torch.Size([4, 8192])loss_mask +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) + torch.Size([4, 8192])batch tensor: +loss_mask batch tensor:torch.Size([4, 8192]) +attention_mask batch tensor:torch.Size([4, 1, 8192, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +attention_mask batch tensor:torch.Size([4, 1, 8192, 8192]) +position_ids batch tensor:torch.Size([4, 8192]) +position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp:batch tensor after cp: tokenslabels torch.Size([4, 2048])torch.Size([4, 2048]) + +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp:batch tensor after cp: loss_masklabels torch.Size([4, 2048])torch.Size([4, 2048]) + +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp:batch tensor after cp: attention_maskloss_mask torch.Size([4, 1, 2048, 8192])torch.Size([4, 2048]) + +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp:batch tensor after cp: position_idsattention_mask torch.Size([4, 2048])torch.Size([4, 1, 2048, 8192]) + +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 6 +Done exporting trace 6 + [2025-06-21 21:27:31] iteration 7/ 10 | consumed samples: 7 | elapsed time per iteration (ms): 71.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 67108864.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048])batch tensor: +batch tensor after cp: loss_mask tokenstorch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192])torch.Size([4, 8192]) + +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_idsbatch tensor: torch.Size([4, 2048])labels + torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048])batch tensor: +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +tokensbatch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 8192])torch.Size([4, 2048]) + +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192])batch tensor after cp: + batch tensor after cp:tokens position_ids torch.Size([4, 2048])torch.Size([4, 2048]) + +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens batch tensor:torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) + batch tensor:tokens labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_masktorch.Size([4, 8192]) +torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor:batch tensor: labelsattention_mask torch.Size([4, 8192])torch.Size([4, 1, 8192, 8192]) + +batch tensor: batch tensor:loss_mask position_idstorch.Size([4, 8192]) +torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp:batch tensor after cp: tokens tokens torch.Size([4, 2048]) +torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: batch tensor after cp:labels labels torch.Size([4, 2048]) +torch.Size([4, 2048]) +batch tensor after cp: batch tensor after cp:loss_mask loss_mask torch.Size([4, 2048])torch.Size([4, 2048]) + +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp:batch tensor after cp: attention_maskattention_mask torch.Size([4, 1, 2048, 8192])torch.Size([4, 1, 2048, 8192]) + +batch tensor after cp:batch tensor after cp: position_idsposition_ids torch.Size([4, 2048])torch.Size([4, 2048]) + +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens batch tensor after cp: tokens torch.Size([4, 2048])torch.Size([4, 8192]) + +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: labelsbatch tensor: torch.Size([4, 2048]) +labelsbatch tensor after cp: loss_masktorch.Size([4, 8192]) +torch.Size([4, 2048])batch tensor: + loss_maskbatch tensor after cp: torch.Size([4, 8192])attention_mask +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) + batch tensor:torch.Size([4, 1, 2048, 8192]) +attention_mask batch tensor after cp: torch.Size([4, 1, 8192, 8192])position_ids +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) + torch.Size([4, 2048])batch tensor: + position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 7 +Done exporting trace 7 + [2025-06-21 21:27:31] iteration 8/ 10 | consumed samples: 8 | elapsed time per iteration (ms): 70.1 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 33554432.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor:batch tensor: loss_mask torch.Size([4, 8192]) +tokensbatch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_idstorch.Size([4, 8192]) torch.Size([4, 8192]) + +batch tensor: labels torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp:batch tensor: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp:batch tensor after cp: attention_masktokens torch.Size([4, 1, 2048, 8192]) +torch.Size([4, 2048])batch tensor after cp: + batch tensor after cp:position_ids labelstorch.Size([4, 2048]) +torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +tokensbatch tensor after cp: position_ids torch.Size([4, 2048]) +torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: batch tensor after cp:tokens tokens torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp:torch.Size([4, 8192]) labels + torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor:batch tensor after cp: labels loss_masktorch.Size([4, 8192]) +torch.Size([4, 2048])batch tensor: +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) + batch tensor after cp:loss_mask attention_masktorch.Size([4, 8192]) +torch.Size([4, 1, 2048, 8192]) +batch tensor:batch tensor after cp: attention_maskposition_ids torch.Size([4, 2048])torch.Size([4, 1, 8192, 8192]) + +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 8 +Done exporting trace 8 + [2025-06-21 21:27:31] iteration 9/ 10 | consumed samples: 9 | elapsed time per iteration (ms): 68.1 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 16777216.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask batch tensor:torch.Size([4, 2048]) +batch tensor after cp: attention_maskbatch tensor: torch.Size([4, 1, 2048, 8192]) + batch tensor after cp: tokensposition_ids torch.Size([4, 2048]) +batch tensor after cp: attention_masktokens torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +torch.Size([4, 8192]) +torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor:batch tensor after cp: loss_mask torch.Size([4, 2048])tokens +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) + batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +torch.Size([4, 8192])batch tensor after cp: + position_ids torch.Size([4, 2048])batch tensor: +batch tensor: position_ids torch.Size([4, 8192]) + labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 9 +Done exporting trace 9 + [2025-06-21 21:27:31] iteration 10/ 10 | consumed samples: 10 | elapsed time per iteration (ms): 68.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 8388608.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +[after training is done] datetime: 2025-06-21 21:27:31 +saving checkpoint at iteration 10 to gpt-checkpoint in torch_dist format +DEBUG:megatron.training.checkpointing:rank: 7, takes 0.07328176498413086 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 6, takes 0.07331705093383789 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 10, takes 0.07404637336730957 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 14, takes 0.07399797439575195 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 9, takes 0.07401752471923828 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 4, takes 0.07334017753601074 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 8, takes 0.07405400276184082 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 2, takes 0.07341146469116211 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 11, takes 0.0741429328918457 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 5, takes 0.07369017601013184 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 15, takes 0.07410717010498047 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 13, takes 0.07415270805358887 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 3, takes 0.0737159252166748 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 12, takes 0.07465958595275879 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 1, takes 0.07377862930297852 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 0, takes 0.31610870361328125 to prepare state dict for ckpt +WARNING:megatron.core.dist_checkpointing.serialization:Overwriting old incomplete / corrupted checkpoint... +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(220612608), 0), (np.int64(221249536), 1), (np.int64(226492416), 2), (np.int64(222298112), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(220612608), 0), (np.int64(221249536), 1), (np.int64(226492416), 2), (np.int64(222298112), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(220612608), 0), (np.int64(221249536), 1), (np.int64(226492416), 2), (np.int64(222298112), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(220612608), 0), (np.int64(221249536), 1), (np.int64(226492416), 2), (np.int64(222298112), 3)] +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.167151689529419 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1607401371002197 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1699011325836182 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.167788028717041 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1608648300170898 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1683316230773926 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1610667705535889 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.167494297027588 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.161147117614746 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 8, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.161620855331421 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 10, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1619641780853271 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1912167072296143 +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 0.010934591293334961 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 11, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.1620213985443115 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 9, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 7, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 4, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 5, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 6, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 2, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 1, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 3, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 0, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 14, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.4072027206420898 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 13, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 15, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 1.4458379745483398 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 12, starting state dict save +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:verifying reuse of global metadata +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:loaded global metadata reuse verification: no loaded plans passed +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 12, plan time: 0.002075672149658203 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.131271 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 13, plan time: 0.04125857353210449 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1313128 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 6.67572021484375e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 14, plan time: 0.050081491470336914 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 3, plan time: 0.27618885040283203 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 1, plan time: 0.2763991355895996 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.131381 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 15, plan time: 0.024245023727416992 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.132686 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1326864 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 2, plan time: 0.27646350860595703 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 8.940696716308594e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.132726 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 5.745887756347656e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 6.103515625e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1314206 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 6.628036499023438e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 4, plan time: 0.27995991706848145 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 7.987022399902344e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 7, plan time: 0.28005266189575195 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1328502 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 7.677078247070312e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 8, plan time: 0.2800562381744385 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1328862 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1315455 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 8.606910705566406e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 7.891654968261719e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.00011944770812988281 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 10, plan time: 0.28012514114379883 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 6, plan time: 0.27841687202453613 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 9, plan time: 0.2796461582183838 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.131752 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1332214 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.131787 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.00011706352233886719 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 7.653236389160156e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 5, plan time: 0.2796294689178467 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 11, plan time: 0.28002047538757324 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.133575 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 8.106231689453125e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.0001583099365234375 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1318796 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 0, plan time: 0.2790968418121338 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 8.392333984375e-05 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750541253.1364937 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.0001163482666015625 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.048355817794799805 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1801014 rank: 13, write(async) time: 0.04878544807434082 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.04836845397949219 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.04762530326843262 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.180705 rank: 11, write(async) time: 0.04882478713989258 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1817496 rank: 5, write(async) time: 0.04817557334899902 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.04884648323059082 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.18104 rank: 10, write(async) time: 0.0492854118347168 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.04962468147277832 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05011320114135742 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.182719 rank: 3, write(async) time: 0.05003046989440918 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.181921 rank: 14, write(async) time: 0.05054116249084473 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.0504913330078125 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05026555061340332 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1823273 rank: 15, write(async) time: 0.05090475082397461 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1834433 rank: 2, write(async) time: 0.05071091651916504 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.051285743713378906 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.051003217697143555 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1829877 rank: 12, write(async) time: 0.05171513557434082 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1841302 rank: 1, write(async) time: 0.05144166946411133 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05146455764770508 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1834543 rank: 8, write(async) time: 0.0519101619720459 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05239582061767578 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1846428 rank: 9, write(async) time: 0.052854061126708984 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05421733856201172 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.187612 rank: 7, write(async) time: 0.05472970008850098 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.054456233978271484 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.188189 rank: 6, write(async) time: 0.05497288703918457 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05552339553833008 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.188937 rank: 4, write(async) time: 0.056084632873535156 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.052221059799194336 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541253.1892061 rank: 0, write(async) time: 0.052709341049194336 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 15, takes 2.3126602172851562e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 13, takes 1.7404556274414062e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 5, takes 1.7881393432617188e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 11, takes 1.5497207641601562e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 10, takes 1.621246337890625e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 8, takes 1.6450881958007812e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 14, takes 1.52587890625e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 9, takes 1.621246337890625e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 12, takes 1.52587890625e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, takes 1.7642974853515625e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 7, takes 1.6927719116210938e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 1, takes 1.9311904907226562e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, takes 1.8358230590820312e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, takes 1.9788742065429688e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, takes 2.6464462280273438e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 15, takes 0.030779600143432617 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 11, takes 0.029303312301635742 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 13, takes 0.031005382537841797 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 5, takes 0.030924558639526367 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 10, takes 0.028961896896362305 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 14, takes 0.029993534088134766 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 8, takes 0.03245258331298828 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 12, takes 0.03090667724609375 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 9, takes 0.03310108184814453 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, takes 1.6927719116210938e-05 to finish D2H +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 7, takes 0.03157830238342285 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, takes 0.030399322509765625 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, takes 0.031244277954101562 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 1, takes 0.03296303749084473 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, takes 0.03704977035522461 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, takes 0.03589940071105957 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, takes 0.03598165512084961 to schedule async ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 8, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 1, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 9, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 10, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 11, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 12, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 5, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 13, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 7, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 14, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 15, joining self.process +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 278528, before: 1749770240, after: 1750048768 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 339968, before: 1736400896, after: 1736740864 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 131072, before: 1751150592, after: 1751281664 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results... +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 17756160, before: 2021625856, after: 2039382016 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114405376, before: 1743781888, after: 1858187264 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114454528, before: 1730334720, after: 1844789248 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114188288, before: 1737154560, after: 1851342848 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114171904, before: 1740169216, after: 1854341120 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 106000384, before: 1738645504, after: 1844645888 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 105701376, before: 1753042944, after: 1858744320 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 106123264, before: 1741475840, after: 1847599104 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 109162496, before: 1743781888, after: 1852944384 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 114192384, before: 1736470528, after: 1850662912 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 109146112, before: 1730334720, after: 1839480832 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 110100480, before: 1751486464, after: 1861586944 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 114130944, before: 1751486464, after: 1865617408 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 114147328, before: 1776353280, after: 1890500608 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 110039040, before: 1776353280, after: 1886392320 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 110215168, before: 1741475840, after: 1851691008 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 118472704, before: 1769848832, after: 1888321536 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 109735936, before: 1738645504, after: 1848381440 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.741687, rank: 5, write(sync,parallel): 0.4366152286529541 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 109289472, before: 1753042944, after: 1862332416 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.749711, rank: 7, write(sync,parallel): 0.4190342426300049 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 117264384, before: 1737154560, after: 1854418944 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 110145536, before: 1736470528, after: 1846616064 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7615688, rank: 11, write(sync,parallel): 0.45766282081604004 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 109072384, before: 1740169216, after: 1849241600 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 122736640, before: 1787523072, after: 1910259712 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7739286, rank: 9, write(sync,parallel): 0.4645233154296875 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114163712, before: 1769848832, after: 1884012544 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.51s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7778873, rank: 14, write(sync,parallel): 0.471665620803833 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.784036, rank: 13, write(sync,parallel): 0.4794635772705078 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.49s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7923906, rank: 15, write(sync,parallel): 0.4897768497467041 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7926157, rank: 10, write(sync,parallel): 0.4881148338317871 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.52s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.7973151, rank: 4, write(sync,parallel): 0.46248936653137207 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 114364416, before: 1787523072, after: 1901887488 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.54s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.54s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.818989, rank: 6, write(sync,parallel): 0.4817230701446533 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.55s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.8238034, rank: 12, write(sync,parallel): 0.5124142169952393 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.56s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.56s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.55s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541253.8487906, rank: 8, write(sync,parallel): 0.5368783473968506 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.57s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.60s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.62s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212434944, before: 1749770240, after: 1962205184 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212430848, before: 1736400896, after: 1948831744 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212160512, before: 2021625856, after: 2233786368 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212492288, before: 1751085056, after: 1963577344 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541254.0919735, rank: 3, write(sync,parallel): 0.7600772380828857 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541254.1060102, rank: 2, write(sync,parallel): 0.7743864059448242 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541254.1234229, rank: 0, write(sync,parallel): 0.7561042308807373 +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully +DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750541254.1315966, rank: 1, write(sync,parallel): 0.7983646392822266 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.84s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.86s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.84s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 0.88s from forking +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1789103, 1, gather: 0.0020093917846679688 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1782258, 9, gather: 0.3667161464691162 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1783838, 10, gather: 0.3503992557525635 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1784444, 12, gather: 0.302262544631958 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1785057, 13, gather: 0.35668420791625977 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1785433, 14, gather: 0.36476969718933105 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1795535, 2, gather: 0.01919412612915039 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1795568, 3, gather: 0.03774881362915039 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1785648, 11, gather: 0.3824899196624756 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.179754, 6, gather: 0.3174264430999756 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1787863, 15, gather: 0.34683918952941895 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1798527, 4, gather: 0.34044742584228516 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1789763, 8, gather: 0.2886927127838135 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1798851, 5, gather: 0.40103626251220703 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1798878, 7, gather: 0.39071202278137207 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1813626, 0, gather: 0.015553712844848633 +DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750541254.1921456, metadata_write: 0.010638236999511719 +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3172s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3713s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3615s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3796s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3032s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3971s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3651s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3816s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0289s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0177s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0526s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3320s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.4155s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.3547s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.4050s +DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0340s + successfully saved checkpoint from iteration 10 to gpt-checkpoint [ t 1/4, p 1/1 ] +DEBUG:megatron.training.checkpointing:rank: 3, takes 0.0021533966064453125 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 1, takes 0.0021893978118896484 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 2, takes 0.002213716506958008 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 4, takes 0.002193927764892578 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 7, takes 0.0021958351135253906 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 15, takes 0.002206087112426758 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 13, takes 0.002221822738647461 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 10, takes 0.002171039581298828 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 12, takes 0.002176046371459961 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 14, takes 0.002187013626098633 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 9, takes 0.0021784305572509766 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 11, takes 0.002147197723388672 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 5, takes 0.002110719680786133 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 6, takes 0.0021288394927978516 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 0, takes 0.002199411392211914 to finalize ckpt save +DEBUG:megatron.training.checkpointing:rank: 8, takes 0.002223491668701172 to finalize ckpt save +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +Evaluating on 1 samples +Evaluating iter 1/1 +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +torch.Size([4, 8192])batch tensor after cp: +labels torch.Size([4, 2048])batch tensor: +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) + batch tensor after cp:labels loss_masktorch.Size([4, 8192]) +torch.Size([4, 2048])batch tensor: + batch tensor after cp:loss_mask attention_masktorch.Size([4, 8192]) +torch.Size([4, 1, 2048, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor:batch tensor after cp: attention_maskposition_ids torch.Size([4, 2048])torch.Size([4, 1, 8192, 8192]) + +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 10 +Done exporting trace 10 +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +(min, max) time across ranks (ms): + evaluate .......................................: (2589.06, 2591.33) +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +---------------------------------------------------------------------------------------------------------------- + validation loss at iteration 10 on validation set | lm loss value: 1.103776E+01 | lm loss PPL: 6.217820E+04 | +Evaluating on 1 samples +Evaluating iter 1/1 +---------------------------------------------------------------------------------------------------------------- +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp:batch tensor: attention_mask tokenstorch.Size([4, 1, 2048, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor: position_ids torch.Size([4, 8192])batch tensor: + tokens torch.Size([4, 8192]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor: labels torch.Size([4, 8192]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: labels torch.Size([4, 2048]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor after cp:batch tensor after cp: loss_masktokens torch.Size([4, 2048]) +torch.Size([4, 2048])batch tensor after cp: + attention_maskbatch tensor after cp: labelstorch.Size([4, 1, 2048, 8192]) +torch.Size([4, 2048])batch tensor after cp: + batch tensor after cp:position_ids loss_masktorch.Size([4, 2048]) +batch tensor: tokens torch.Size([4, 8192]) +batch tensor: labels torch.Size([4, 8192]) +torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor: loss_mask torch.Size([4, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +batch tensor: attention_mask torch.Size([4, 1, 8192, 8192]) +batch tensor: position_ids torch.Size([4, 8192]) +(min, max) time across ranks (ms): + evaluate .......................................: (45.15, 46.90) +batch tensor after cp: tokens torch.Size([4, 2048]) +batch tensor after cp: labels torch.Size([4, 2048]) +---------------------------------------------------------------------------------------------------------- + validation loss at iteration 10 on test set | lm loss value: 1.103776E+01 | lm loss PPL: 6.217820E+04 | +---------------------------------------------------------------------------------------------------------- +batch tensor after cp: loss_mask torch.Size([4, 2048]) +batch tensor after cp: attention_mask torch.Size([4, 1, 2048, 8192]) +batch tensor after cp: position_ids torch.Size([4, 2048]) +Start exporting trace 11 +Done exporting trace 11 +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +WARNING:megatron.core.rerun_state_machine:Setting RerunStateMachine mode RerunMode.DISABLED +Running ctx_length=4096, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 4096 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +-------------------------------- +CTX_LENGTH: 4096 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 4096 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 4096 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 4096 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.050 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.627 seconds +time to initialize megatron (seconds): 8.273 +[after megatron is initialized] datetime: 2025-06-21 21:28:15 +building GPT model ... +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 156830720 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 156830720 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 156830720 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (156830720 elements, 156830720 padded size): + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.embedding.word_embeddings.weight + module.decoder.final_layernorm.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.bias + module.embedding.position_embeddings.weight + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.self_attention.linear_proj.weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_proj.bias + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.bias +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine + loading distributed checkpoint from gpt-checkpoint at iteration 10 +Running ctx_length=8192, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 8192 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 8192 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 8192 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 8192 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 8192 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.053 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.433 seconds +time to initialize megatron (seconds): 8.256 +[after megatron is initialized] datetime: 2025-06-21 21:28:53 +building GPT model ... +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 173607936 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 173607936 +>>> embedding>>> embedding + +>>> decoder>>> decoder + +>>> output_layer>>> output_layer + +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 173607936 > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 173607936 + +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 173607936 + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 173607936 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 173607936 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 173607936 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 173607936 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (173607936 elements, 173607936 padded size): + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.self_attention.linear_qkv.bias + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.embedding.word_embeddings.weight + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.decoder.final_layernorm.weight + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.layers.0.self_attention.linear_proj.weight + module.embedding.position_embeddings.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_proj.bias + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.bias +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine + loading distributed checkpoint from gpt-checkpoint at iteration 10 +Running ctx_length=12288, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 12288 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 12288 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 12288 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 12288 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 12288 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.054 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.580 seconds +time to initialize megatron (seconds): 8.637 +[after megatron is initialized] datetime: 2025-06-21 21:29:31 +building GPT model ... +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 190385152 + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 190385152 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 190385152 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 190385152 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (190385152 elements, 190385152 padded size): + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.embedding.word_embeddings.weight + module.decoder.final_layernorm.weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.bias + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.embedding.position_embeddings.weight + module.decoder.layers.0.mlp.linear_fc1.bias + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.layers.0.self_attention.linear_proj.weight + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.decoder.layers.0.self_attention.linear_proj.bias +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine +WARNING: could not find the metadata file gpt-checkpoint/latest_checkpointed_iteration.txt + will not load any checkpoints and will start from random +(min, max) time across ranks (ms): + load-checkpoint ................................: (3.24, 3.49) +[after model, optimizer, and learning rate scheduler are built] datetime: 2025-06-21 21:29:32 +> building train, validation, and test datasets ... + > datasets target sizes (minimum size): + train: 10 + validation: 1 + test: 1 +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let mock = True, as both blend and blend_per_split are None +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split = 1,1,1, an arbitrarily even split, as mock is True +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 0.3333333333333333), (0.3333333333333333, 0.6666666666666666), (0.6666666666666666, 1.0)] +> building train, validation, and test datasets for GPT ... +INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building MockGPTDataset splits with sizes=(10, 1, 1) and config=GPTDatasetConfig(random_seed=1234, sequence_length=12288, blend=None, blend_per_split=None, split='1,1,1', split_matrix=[(0, 0.3333333333333333), (0.3333333333333333, 0.6666666666666666), (0.6666666666666666, 1.0)], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=True, tokenizer=, mid_level_dataset_surplus=0.005, reset_position_ids=False, reset_attention_mask=False, eod_mask_loss=False, create_attention_mask=True, drop_last_partial_validation_sequence=True, add_extra_token_to_sequence=True, object_storage_cache_path=None) +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset train indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.005003 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 5549 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset valid indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.001853 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 5546 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +INFO:megatron.core.datasets.gpt_dataset:Build and save the MockGPTDataset test indices +DEBUG:megatron.core.datasets.gpt_dataset:> separate_final_epoch: False +WARNING:megatron.core.datasets.gpt_dataset:Unable to save MockGPTDataset indexes because path_to_cache is None +DEBUG:megatron.core.datasets.gpt_dataset: > time elapsed: 0.001734 seconds +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 5557 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +> finished creating GPT datasets ... +[after dataloaders are built] datetime: 2025-06-21 21:29:32 +done with setup ... +training ... +(min, max) time across ranks (ms): + model-and-optimizer-setup ......................: (885.41, 914.23) + train/valid/test-data-iterators-setup ..........: (16.92, 143.18) +Setting rerun_state_machine.current_iteration to 0... +[before the start of training step] datetime: 2025-06-21 21:29:32 +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 0 +Done exporting trace 0 +Number of parameters in transformer block in billions: 0.35 +Number of parameters in embedding layers in billions: 0.21 +Total number of parameters in billions: 0.56 +Number of parameters in most loaded shard in billions: 0.1400 +Theoretical memory footprints: weight and optimizer=2403.18 MB + [2025-06-21 21:29:47] iteration 1/ 10 | consumed samples: 1 | elapsed time per iteration (ms): 15050.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 4294967296.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +[Rank 1] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24820.0 | max reserved: 24820.0 +[Rank 9] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25360.0 | max reserved: 25360.0 +[Rank 11] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 24976.0 | max reserved: 24976.0 +[Rank 3] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24820.0 | max reserved: 24820.0[Rank 0] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24820.0 | max reserved: 24820.0 + +[Rank 8] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25360.0 | max reserved: 25360.0 +[Rank 10] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25120.0 | max reserved: 25120.0 +[Rank 5] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24964.0 | max reserved: 24964.0 +[Rank 12] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25156.0 | max reserved: 25156.0 +[Rank 6] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24868.0 | max reserved: 24868.0 +[Rank 4] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24868.0 | max reserved: 24868.0 +[Rank 7] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 24868.0 | max reserved: 24868.0 +[Rank 14] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25156.0 | max reserved: 25156.0 +[Rank 2] (after 1 iterations) memory (MB) | allocated: 13856.22314453125 | max allocated: 23434.99658203125 | reserved: 25204.0 | max reserved: 25204.0 +[Rank 13] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25156.0 | max reserved: 25156.0 +[Rank 15] (after 1 iterations) memory (MB) | allocated: 13855.22314453125 | max allocated: 23434.99658203125 | reserved: 25156.0 | max reserved: 25156.0 +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 1 +Done exporting trace 1 + [2025-06-21 21:29:49] iteration 2/ 10 | consumed samples: 2 | elapsed time per iteration (ms): 1758.7 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 2147483648.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 2 +Done exporting trace 2 + [2025-06-21 21:29:51] iteration 3/ 10 | consumed samples: 3 | elapsed time per iteration (ms): 1766.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 1073741824.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 3 +Done exporting trace 3 + [2025-06-21 21:29:52] iteration 4/ 10 | consumed samples: 4 | elapsed time per iteration (ms): 1760.8 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 536870912.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 4 +Done exporting trace 4 + [2025-06-21 21:29:54] iteration 5/ 10 | consumed samples: 5 | elapsed time per iteration (ms): 1769.8 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 268435456.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 5 +Done exporting trace 5 + [2025-06-21 21:29:56] iteration 6/ 10 | consumed samples: 6 | elapsed time per iteration (ms): 1769.9 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 134217728.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 49152])torch.Size([4, 12288]) + +batch tensor after cp: loss_maskbatch tensor: torch.Size([4, 12288])labels + batch tensor after cp: torch.Size([4, 49152])attention_mask + batch tensor:torch.Size([4, 1, 12288, 49152]) +loss_mask batch tensor after cp: torch.Size([4, 49152])position_ids + torch.Size([4, 12288]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: batch tensor after cp:labels tokens torch.Size([4, 12288]) +torch.Size([4, 12288])batch tensor after cp: + batch tensor after cp:loss_mask labelstorch.Size([4, 12288]) +torch.Size([4, 12288]) +batch tensor after cp: batch tensor after cp:attention_mask loss_masktorch.Size([4, 1, 12288, 49152]) +torch.Size([4, 12288])batch tensor after cp: + batch tensor after cp:position_ids torch.Size([4, 12288])attention_mask +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) + torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 6 +Done exporting trace 6 + [2025-06-21 21:29:58] iteration 7/ 10 | consumed samples: 7 | elapsed time per iteration (ms): 1756.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 67108864.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 7 +Done exporting trace 7 + [2025-06-21 21:29:59] iteration 8/ 10 | consumed samples: 8 | elapsed time per iteration (ms): 1754.8 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 33554432.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 8 +Done exporting trace 8 + [2025-06-21 21:30:01] iteration 9/ 10 | consumed samples: 9 | elapsed time per iteration (ms): 1754.7 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 16777216.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +batch tensor: tokens torch.Size([4, 49152]) +batch tensor: labels torch.Size([4, 49152]) +batch tensor: loss_mask torch.Size([4, 49152]) +batch tensor: attention_mask torch.Size([4, 1, 49152, 49152]) +batch tensor: position_ids torch.Size([4, 49152]) +batch tensor after cp: tokens torch.Size([4, 12288]) +batch tensor after cp: labels torch.Size([4, 12288]) +batch tensor after cp: loss_mask torch.Size([4, 12288]) +batch tensor after cp: attention_mask torch.Size([4, 1, 12288, 49152]) +batch tensor after cp: position_ids torch.Size([4, 12288]) +Start exporting trace 9 +Done exporting trace 9 + [2025-06-21 21:30:04] iteration 10/ 10 | consumed samples: 10 | elapsed time per iteration (ms): 2901.0 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 8388608.0 | number of skipped iterations: 1 | number of nan iterations: 0 | +[after training is done] datetime: 2025-06-21 21:30:04 +saving checkpoint at iteration 10 to gpt-checkpoint in torch_dist format +DEBUG:megatron.training.checkpointing:rank: 5, takes 0.028438329696655273 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 7, takes 0.02845907211303711 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 6, takes 0.028488636016845703 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 3, takes 0.028661012649536133 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 1, takes 0.02868056297302246 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 14, takes 0.029039859771728516 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 4, takes 0.02912139892578125 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 11, takes 0.029221057891845703 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 10, takes 0.029230356216430664 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 13, takes 0.02927422523498535 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 0, takes 0.029855012893676758 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 8, takes 0.02971816062927246 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 15, takes 0.030900239944458008 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 12, takes 0.03170156478881836 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 9, takes 0.040781497955322266 to prepare state dict for ckpt +DEBUG:megatron.training.checkpointing:rank: 2, takes 0.04349184036254883 to prepare state dict for ckpt +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:Apply save parallelization +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(212860928), 1), (np.int64(213909504), 2), (np.int64(205588480), 3)] +Running ctx_length=16384, TP_SIZE=4, CP_SIZE=4, BATCH_SIZE=4 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 16384 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Cleaning up checkpoint directory: gpt-checkpoint +-------------------------------- +CTX_LENGTH: 16384 +TP_SIZE: 4 +CP_SIZE: 4 +CHECKPOINT_PATH: gpt-checkpoint +PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron +-------------------------------- +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +using world size: 16, data-parallel size: 1, context-parallel size: 4, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 4, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0 +Number of virtual stages per pipeline stage: None +WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used +using torch.float16 for parameters ... +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. False + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. True + add_position_embedding .......................... True + add_qkv_bias .................................... True + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... False + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.1 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ False + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ True + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 4 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 1 + data_path ....................................... None + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + deprecated_use_mcore_models ..................... False + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 2 + encoder_pipeline_model_parallel_size ............ 0 + encoder_seq_length .............................. 16384 + encoder_tensor_model_parallel_size .............. 0 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 16 + eval_iters ...................................... 1 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 1 + expert_tensor_parallel_size ..................... 4 + external_cuda_graph ............................. False + ffn_hidden_size ................................. 16384 + finetune ........................................ False + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ True + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. None + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... delayed + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 1 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.1 + hidden_size ..................................... 4096 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.02 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 16 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kv_channels ..................................... 64 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ gpt-checkpoint + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. False + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... 0 + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0005 + lr_decay_iters .................................. 150000 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 2 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... True + max_position_embeddings ......................... 16384 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 0.0 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... True + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.0 + moe_enable_deepep ............................... False + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. None + moe_grouped_gemm ................................ False + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ None + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 2 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... allgather + moe_token_drop_policy ........................... probs + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 1.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-05 + normalization ................................... LayerNorm + num_attention_heads ............................. 64 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... None + num_layers ...................................... 2 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 16 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ False + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.float16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_size .................... 1 + pipeline_model_parallel_split_rank .............. None + position_embedding_type ......................... learned_absolute + pretrained_checkpoint ........................... None + profile ......................................... False + profile_ranks ................................... [0] + profile_step_end ................................ 12 + profile_step_start .............................. 10 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... False + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... disabled + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + rope_scaling_factor ............................. 8.0 + rotary_base ..................................... 10000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ gpt-checkpoint + save_interval ................................... 16 + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 16384 + sequence_parallel ............................... False + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... None + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... False + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. False + tensor_model_parallel_size ...................... 4 + tensorboard_dir ................................. tensorboard-logs/ + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. False + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... False + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 16 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 + > padded vocab (size: 50257) with 431 dummy tokens (new size: 50688) +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED +> initializing torch distributed ... +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written. +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +INFO:megatron.training.initialize:Setting logging level to 0 +INFO:megatron.training.initialize:Setting logging level to 0 +> initialized tensor model parallel with size 4 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +> compiling dataset index builder ... +make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.040 seconds +> compiling and loading fused kernels ... +>>> done with compiling and loading fused kernels. Compilation time: 2.360 seconds +time to initialize megatron (seconds): 8.241 +[after megatron is initialized] datetime: 2025-06-21 21:31:21 +building GPT model ... +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 207162368 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 207162368 + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 207162368 + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 207162368 +>>> embedding +>>> decoder +>>> output_layer + > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 207162368 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=False, overlap_grad_reduce=False, overlap_param_gather=False, align_param_gather=False, use_distributed_optimizer=False, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=None, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 1 +Params for bucket 1 (207162368 elements, 207162368 padded size): + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.final_layernorm.bias + module.decoder.layers.1.mlp.linear_fc2.weight + module.decoder.layers.1.self_attention.linear_proj.bias + module.decoder.layers.1.mlp.linear_fc1.layer_norm_bias + module.decoder.layers.0.mlp.linear_fc2.weight + module.decoder.layers.0.mlp.linear_fc1.layer_norm_bias + module.decoder.final_layernorm.weight + module.decoder.layers.1.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.1.self_attention.linear_qkv.bias + module.decoder.layers.0.mlp.linear_fc2.bias + module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight + module.decoder.layers.0.self_attention.linear_qkv.bias + module.decoder.layers.1.mlp.linear_fc1.weight + module.decoder.layers.0.mlp.linear_fc1.weight + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.1.mlp.linear_fc2.bias + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.self_attention.linear_proj.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_bias + module.decoder.layers.0.self_attention.linear_proj.bias + module.embedding.position_embeddings.weight + module.embedding.word_embeddings.weight + module.decoder.layers.1.mlp.linear_fc1.bias + module.decoder.layers.0.mlp.linear_fc1.bias +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0005, min_lr=0.0, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp16=True, bf16=False, params_dtype=torch.float16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=False, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='')