madlag's picture
Adding model, graphs and metadata.
c0854f5
{
"checkpoint_path": "/data_2to/devel_data/nn_pruning/output/squad_test_9_fullpatch4/hp_od-__data_2to__devel_data__nn_pruning__output__squad_test_9_fullpatch4___es-steps_nte20_ls250_stl50_est5000_rn-__data_2to__devel_data__nn_pruning__output__squad_test_9_fullpatch4_--6cb2db64e9a885f1/checkpoint-110000",
"config": {
"_name_or_path": "/tmp/tmptzy6fw4l",
"architectures": ["BertForQuestionAnswering"],
"attention_probs_dropout_prob": 0.1,
"gradient_checkpointing": false,
"hidden_act": "relu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"layer_norm_type": "no_norm",
"max_position_embeddings": 512,
"model_type": "bert",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"pad_token_id": 0,
"position_embedding_type": "absolute",
"pruned_heads": {
"0": [0, 2, 4, 5, 6],
"1": [0, 2, 3, 5, 6, 7, 8],
"10": [1, 2, 4, 5, 6, 7, 8],
"11": [0, 5, 7, 8, 10, 11],
"2": [4, 7, 8],
"3": [2, 4, 6],
"4": [1, 2],
"5": [1, 2, 6, 7, 11],
"6": [2, 3, 10],
"7": [1, 3, 6, 7, 11],
"8": [0, 3, 4],
"9": [1, 4, 5, 7, 9, 10]
},
"type_vocab_size": 2,
"vocab_size": 30522
},
"eval_metrics": {
"exact_match": 81.30558183538317,
"f1": 88.32616013981377,
"main_metric": 88.32616013981377
},
"model_args": {
"cache_dir": null,
"config_name": null,
"model_name_or_path": "bert-base-uncased",
"tokenizer_name": null,
"use_fast_tokenizer": true
},
"sparse_args": {
"ampere_pruning_method": "disabled",
"attention_block_cols": 32,
"attention_block_rows": 32,
"attention_lambda": 1.0,
"attention_output_with_dense": 0,
"attention_pruning_method": "sigmoied_threshold",
"bias_mask": true,
"dense_block_cols": 1,
"dense_block_rows": 1,
"dense_lambda": 1.0,
"dense_pruning_method": "sigmoied_threshold:1d_alt",
"distil_alpha_ce": 0.1,
"distil_alpha_teacher": 0.9,
"distil_teacher_name_or_path": "bert-large-uncased-whole-word-masking-finetuned-squad",
"distil_temperature": 2.0,
"final_ampere_temperature": 20.0,
"final_finetune": false,
"final_threshold": 0.1,
"final_warmup": 10,
"gelu_patch": 1,
"gelu_patch_steps": 50000,
"initial_ampere_temperature": 0.0,
"initial_threshold": 0,
"initial_warmup": 1,
"layer_norm_patch": 1,
"layer_norm_patch_start_delta": 0.99,
"layer_norm_patch_steps": 50000,
"linear_min_parameters": 0,
"mask_init": "constant",
"mask_scale": 0.0,
"mask_scores_learning_rate": 0.01,
"regularization": "l1",
"regularization_final_lambda": 10
},
"speed": {
"cuda_eval_elapsed_time": 19.66401289367676,
"eval_elapsed_time": 26.647668646648526
},
"speedup": 1.9626916038984936,
"stats": {
"layers": {
"0": {
"linear_attention_nnz": 920576,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1188864,
"linear_dense_total": 4718592,
"linear_nnz": 2109440,
"linear_total": 7077888,
"nnz": 2115942,
"total": 7087872
},
"1": {
"linear_attention_nnz": 709632,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1187328,
"linear_dense_total": 4718592,
"linear_nnz": 1896960,
"linear_total": 7077888,
"nnz": 1903205,
"total": 7087872
},
"10": {
"linear_attention_nnz": 617472,
"linear_attention_total": 2359296,
"linear_dense_nnz": 552960,
"linear_dense_total": 4718592,
"linear_nnz": 1170432,
"linear_total": 7077888,
"nnz": 1176200,
"total": 7087872
},
"11": {
"linear_attention_nnz": 423936,
"linear_attention_total": 2359296,
"linear_dense_nnz": 806400,
"linear_dense_total": 4718592,
"linear_nnz": 1230336,
"linear_total": 7077888,
"nnz": 1236333,
"total": 7087872
},
"2": {
"linear_attention_nnz": 1210368,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1371648,
"linear_dense_total": 4718592,
"linear_nnz": 2582016,
"linear_total": 7077888,
"nnz": 2589021,
"total": 7087872
},
"3": {
"linear_attention_nnz": 1433600,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1310208,
"linear_dense_total": 4718592,
"linear_nnz": 2743808,
"linear_total": 7077888,
"nnz": 2750901,
"total": 7087872
},
"4": {
"linear_attention_nnz": 1558528,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1219584,
"linear_dense_total": 4718592,
"linear_nnz": 2778112,
"linear_total": 7077888,
"nnz": 2785338,
"total": 7087872
},
"5": {
"linear_attention_nnz": 1038336,
"linear_attention_total": 2359296,
"linear_dense_nnz": 1213440,
"linear_dense_total": 4718592,
"linear_nnz": 2251776,
"linear_total": 7077888,
"nnz": 2258486,
"total": 7087872
},
"6": {
"linear_attention_nnz": 1107968,
"linear_attention_total": 2359296,
"linear_dense_nnz": 967680,
"linear_dense_total": 4718592,
"linear_nnz": 2075648,
"linear_total": 7077888,
"nnz": 2082422,
"total": 7087872
},
"7": {
"linear_attention_nnz": 1108992,
"linear_attention_total": 2359296,
"linear_dense_nnz": 766464,
"linear_dense_total": 4718592,
"linear_nnz": 1875456,
"linear_total": 7077888,
"nnz": 1881875,
"total": 7087872
},
"8": {
"linear_attention_nnz": 1062912,
"linear_attention_total": 2359296,
"linear_dense_nnz": 399360,
"linear_dense_total": 4718592,
"linear_nnz": 1462272,
"linear_total": 7077888,
"nnz": 1468548,
"total": 7087872
},
"9": {
"linear_attention_nnz": 670720,
"linear_attention_total": 2359296,
"linear_dense_nnz": 274944,
"linear_dense_total": 4718592,
"linear_nnz": 945664,
"linear_total": 7077888,
"nnz": 951443,
"total": 7087872
}
},
"linear_nnz": 23121920,
"linear_sparsity": 72.77681327160495,
"linear_total": 84934656,
"nnz": 47038436,
"pruned_heads": {
"0": [0, 2, 4, 5, 6],
"1": [0, 2, 3, 5, 6, 7, 8],
"10": [1, 2, 4, 5, 6, 7, 8],
"11": [0, 5, 7, 8, 10, 11],
"2": [8, 4, 7],
"3": [2, 4, 6],
"4": [1, 2],
"5": [1, 2, 6, 7, 11],
"6": [10, 2, 3],
"7": [1, 3, 6, 7, 11],
"8": [0, 3, 4],
"9": [1, 4, 5, 7, 9, 10]
},
"total": 108893186,
"total_sparsity": 56.8031410156371
},
"training_args": {
"_n_gpu": -1,
"adafactor": false,
"adam_beta1": 0.9,
"adam_beta2": 0.999,
"adam_epsilon": 1e-08,
"dataloader_drop_last": false,
"dataloader_num_workers": 0,
"dataloader_pin_memory": true,
"ddp_find_unused_parameters": null,
"debug": false,
"deepspeed": null,
"disable_tqdm": false,
"do_eval": 1,
"do_predict": false,
"do_train": 1,
"eval_accumulation_steps": null,
"eval_steps": 5000,
"evaluation_strategy": "steps",
"fp16": false,
"fp16_backend": "auto",
"fp16_full_eval": false,
"fp16_opt_level": "O1",
"gradient_accumulation_steps": 1,
"greater_is_better": null,
"group_by_length": false,
"ignore_data_skip": false,
"label_names": null,
"label_smoothing_factor": 0.0,
"learning_rate": 3e-05,
"length_column_name": "length",
"load_best_model_at_end": false,
"local_rank": -1,
"logging_dir": "/data_2to/devel_data/nn_pruning/output/squad_test_9_fullpatch4/",
"logging_first_step": false,
"logging_steps": 250,
"logging_strategy": "steps",
"lr_scheduler_type": "linear",
"max_grad_norm": 1.0,
"max_steps": -1,
"metric_for_best_model": null,
"mp_parameters": "",
"no_cuda": false,
"num_train_epochs": 20,
"optimize_model_before_eval": "disabled",
"output_dir": "/data_2to/devel_data/nn_pruning/output/squad_test_9_fullpatch4/",
"overwrite_output_dir": 1,
"past_index": -1,
"per_device_eval_batch_size": 8,
"per_device_train_batch_size": 16,
"per_gpu_eval_batch_size": null,
"per_gpu_train_batch_size": null,
"prediction_loss_only": false,
"remove_unused_columns": true,
"report_to": null,
"run_name": "/data_2to/devel_data/nn_pruning/output/squad_test_9_fullpatch4/",
"save_steps": 5000,
"save_strategy": "steps",
"save_total_limit": 50,
"seed": 17,
"sharded_ddp": "",
"skip_memory_metrics": false,
"tpu_metrics_debug": false,
"tpu_num_cores": null,
"warmup_ratio": 0.0,
"warmup_steps": 5400,
"weight_decay": 0.0
}
}