Upload folder using huggingface_hub

.gitattributes

@@ -63,3 +63,4 @@ attnserver.run_attnserver.slurm.sh.343192.err.log filter=lfs diff=lfs merge=lfs
 attnserver.run_attnserver.slurm.sh.343194.err.log filter=lfs diff=lfs merge=lfs -text
 attnserver.run_attnserver.slurm.sh.343196.err.log filter=lfs diff=lfs merge=lfs -text
 attnserver.run_attnserver.slurm.sh.343205.err.log filter=lfs diff=lfs merge=lfs -text
+attnserver.run_attnserver.slurm.sh.343215.err.log filter=lfs diff=lfs merge=lfs -text

attnserver.run_attnserver.slurm.sh.343207.out.log

@@ -19372,3 +19372,252 @@ batch tensor after cp: position_ids torch.Size([1, 131072])
 Start exporting trace 1
 Done exporting trace 1
  [2025-06-21 21:59:22] iteration        2/      10 | consumed samples:            2 | elapsed time per iteration (ms): 41689.2 | 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([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+Start exporting trace 2
+Done exporting trace 2
+ [2025-06-21 22:02:29] iteration        3/      10 | consumed samples:            3 | elapsed time per iteration (ms): 187533.7 | 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([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+Start exporting trace 3
+Done exporting trace 3
+ [2025-06-21 22:04:53] iteration        4/      10 | consumed samples:            4 | elapsed time per iteration (ms): 144038.9 | 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([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 131072])
+batch tensor after cp: labels torch.Size([1, 131072])
+batch tensor after cp: loss_mask torch.Size([1, 131072])
+batch tensor after cp: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor after cp: position_ids torch.Size([1, 131072])
+Start exporting trace 4
+Done exporting trace 4
+ [2025-06-21 22:06:58] iteration        5/      10 | consumed samples:            5 | elapsed time per iteration (ms): 124682.2 | learning rate: 0.000000E+00 | global batch size:     1 | loss scale: 268435456.0 | number of skipped iterations:   1 | number of nan iterations:   0 |

attnserver.run_attnserver.slurm.sh.343215.out.log

@@ -4148,3 +4148,40 @@ DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks
 DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(103809024), 1), (np.int64(92274688), 2), (np.int64(92274688), 3), (np.int64(83919872), 4), (np.int64(83919872), 5), (np.int64(88080384), 6), (np.int64(88080384), 7)]
 DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(103809024), 1), (np.int64(92274688), 2), (np.int64(92274688), 3), (np.int64(83919872), 4), (np.int64(83919872), 5), (np.int64(88080384), 6), (np.int64(88080384), 7)]
 DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(207618048), 0), (np.int64(103809024), 1), (np.int64(92274688), 2), (np.int64(92274688), 3), (np.int64(83919872), 4), (np.int64(83919872), 5), (np.int64(88080384), 6), (np.int64(88080384), 7)]
+Running ctx_length=2048, TP_SIZE=4, CP_SIZE=8, BATCH_SIZE=4
+Cleaning up checkpoint directory: gpt-checkpoint
+--------------------------------
+CTX_LENGTH: 2048
+TP_SIZE: 4
+CP_SIZE: 8
+CHECKPOINT_PATH: gpt-checkpoint
+PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
+--------------------------------
+Cleaning up checkpoint directory: gpt-checkpoint
+/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: 8
+CHECKPOINT_PATH: gpt-checkpoint
+PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
+--------------------------------
+CTX_LENGTH: 2048
+TP_SIZE: 4
+CP_SIZE: 8
+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
+Cleaning up checkpoint directory: gpt-checkpoint
+--------------------------------
+CTX_LENGTH: 2048
+TP_SIZE: 4
+CP_SIZE: 8
+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

attnserver.run_attnserver.slurm.sh.343225.err.log

@@ -2613,3 +2613,146 @@ W0621 21:51:26.238000 2242503 site-packages/torch/distributed/run.py:766] ******
   warnings.warn(
 /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
   warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank0]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank0]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank0]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank0]:     finalize_fn()
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank0]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 243, in save_state_dict_async_finalize
+[rank0]:     storage_writer.finish(global_metadata, all_results)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 483, in finish
+[rank0]:     super().finish(metadata, results)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/checkpoint/filesystem.py", line 697, in finish
+[rank0]:     with self.fs.create_stream(tmp_path, "wb") as metadata_file:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/contextlib.py", line 137, in __enter__
+[rank0]:     return next(self.gen)
+[rank0]:            ^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/checkpoint/filesystem.py", line 476, in create_stream
+[rank0]:     with path.open(mode) as stream:
+[rank0]:          ^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/pathlib.py", line 1013, in open
+[rank0]:     return io.open(self, mode, buffering, encoding, errors, newline)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/.metadata.tmp'
+[rank0]:[W621 22:04:56.500419579 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:05:35.880000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242575 closing signal SIGTERM
+W0621 22:05:35.885000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242576 closing signal SIGTERM
+W0621 22:05:35.888000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242577 closing signal SIGTERM
+W0621 22:05:35.891000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242578 closing signal SIGTERM
+W0621 22:05:35.892000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242579 closing signal SIGTERM
+W0621 22:05:35.904000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242580 closing signal SIGTERM
+W0621 22:05:35.929000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2242581 closing signal SIGTERM
+E0621 22:06:00.973000 2242503 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 2242574) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:05:35
+  host      : fs-mbz-gpu-768
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 2242574)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=131072
++ PROF_CTX_LENGTH=131072
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L131072*tp4.cp2.bs1.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L131072*tp4.cp2.bs1.json' ']'
++ echo 'Running ctx_length=131072, TP_SIZE=4, CP_SIZE=2, BATCH_SIZE=1'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343225 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-768:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 4 --context-parallel-size 2 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 131072 --max-position-embeddings 131072 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:06:05.199000 2247188 site-packages/torch/distributed/run.py:766] 
+W0621 22:06:05.199000 2247188 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:06:05.199000 2247188 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:06:05.199000 2247188 site-packages/torch/distributed/run.py:766] *****************************************
+[rank6]:[W621 22:06:27.689693214 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:06:27.689715067 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:06:27.695981890 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:06:27.696144469 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:06:27.696472936 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:06:27.697356152 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank1]:[W621 22:06:27.697670162 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:06:27.853656033 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(

attnserver.run_attnserver.slurm.sh.343225.out.log

@@ -21443,3 +21443,1188 @@ batch tensor after cp: labels torch.Size([1, 49152])
 batch tensor after cp: loss_mask torch.Size([1, 49152])
 batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
 batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+Start exporting trace 6
+Done exporting trace 6
+ [2025-06-21 21:59:39] iteration        7/      10 | consumed samples:            7 | elapsed time per iteration (ms): 75055.9 | 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([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+Start exporting trace 7
+Done exporting trace 7
+ [2025-06-21 22:01:00] iteration        8/      10 | consumed samples:            8 | elapsed time per iteration (ms): 80849.2 | 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([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+Start exporting trace 8
+Done exporting trace 8
+ [2025-06-21 22:02:04] iteration        9/      10 | consumed samples:            9 | elapsed time per iteration (ms): 64767.7 | 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([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+batch tensor: tokens torch.Size([1, 98304])
+batch tensor: labels torch.Size([1, 98304])
+batch tensor: loss_mask torch.Size([1, 98304])
+batch tensor: attention_mask torch.Size([1, 1, 98304, 98304])
+batch tensor: position_ids torch.Size([1, 98304])
+batch tensor after cp: tokens torch.Size([1, 49152])
+batch tensor after cp: labels torch.Size([1, 49152])
+batch tensor after cp: loss_mask torch.Size([1, 49152])
+batch tensor after cp: attention_mask torch.Size([1, 1, 49152, 98304])
+batch tensor after cp: position_ids torch.Size([1, 49152])
+Start exporting trace 9
+Done exporting trace 9
+ [2025-06-21 22:03:38] iteration       10/      10 | consumed samples:           10 | elapsed time per iteration (ms): 93378.6 | 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 22:03:38 
+saving checkpoint at iteration      10 to gpt-checkpoint in torch_dist format
+DEBUG:megatron.training.checkpointing:rank: 2, takes 0.07908964157104492 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 6, takes 0.07910752296447754 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 1, takes 0.07910609245300293 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 5, takes 0.07912921905517578 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 7, takes 0.07905840873718262 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 3, takes 0.07906150817871094 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 0, takes 0.0825493335723877 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 4, takes 0.14761972427368164 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.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(417400832), 0), (np.int64(422576128), 1)]
+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(1631584256), 0), (np.int64(1624655872), 1)]
+DEBUG:megatron.core.dist_checkpointing.exchange_utils:distribute_shards_to_ranks distribution: [(np.int64(1631584256), 0), (np.int64(1624655872), 1)]
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.76030206680298
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.76056408882141
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.760433197021484
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.760650873184204
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.760414123535156
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 37.76091694831848
+DEBUG:megatron.core.dist_checkpointing.strategies.fully_parallel:parallel save sharding, time: 0.028193950653076172
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 6, starting state dict save
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 5, 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: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: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: 0, starting state dict save
+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: 3, 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: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: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: 37.775845527648926
+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:loaded global metadata reuse verification: no loaded plans passed
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 5, plan time: 0.07368707656860352
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2063708
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 6, plan time: 0.07391166687011719
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2064087
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 7, plan time: 0.07307553291320801
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2065277
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 1, plan time: 0.06815838813781738
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 2, plan time: 0.06845450401306152
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2066731
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2066875
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 9.5367431640625e-05
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 3, plan time: 0.0684041976928711
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.206915
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.00032067298889160156
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 9.72747802734375e-05
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.0005800724029541016
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.0006380081176757812
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.0005955696105957031
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 4, plan time: 0.02703571319580078
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.210689
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 0.001257181167602539
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:rank: 0, plan time: 0.0778207778930664
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:thread_count: 2, time: 1750543457.2154331
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:bucket_prep, time: 6.413459777832031e-05
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05709362030029297
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05720639228820801
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.2645063 rank: 6, write(async) time: 0.05809736251831055
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05711174011230469
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.2645593 rank: 3, write(async) time: 0.057645559310913086
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.2646434 rank: 7, write(async) time: 0.05811643600463867
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.057874202728271484
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05771374702453613
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.264992 rank: 2, write(async) time: 0.05830240249633789
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.265057 rank: 1, write(async) time: 0.05838346481323242
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.05812430381774902
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.265517 rank: 5, write(async) time: 0.05914449691772461
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.07155466079711914
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.2875042 rank: 0, write(async) time: 0.07206296920776367
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:D2H and push, time: 0.1144871711730957
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543457.3269794 rank: 4, write(async) time: 0.11631417274475098
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 7, takes 1.8358230590820312e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, takes 1.811981201171875e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 5, takes 1.8835067749023438e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 7, takes 0.0258786678314209 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, takes 0.025389671325683594 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 5, takes 0.026950359344482422 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, takes 1.8358230590820312e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, 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.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:1 started
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collecting worker results...
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, takes 0.02531123161315918 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, takes 0.025214672088623047 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 1, takes 0.025298357009887695 to schedule async ckpt 
+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: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:FileSystemWriterAsync: collecting worker results...
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 started
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 217210880, before: 1648107520, after: 1865318400
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, takes 1.9311904907226562e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 218206208, before: 1620013056, after: 1838219264
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 218083328, before: 1636126720, after: 1854210048
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, takes 0.038712263107299805 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 214204416, before: 1620013056, after: 1834217472
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 214429696, before: 1614884864, after: 1829314560
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 214933504, before: 1636126720, after: 1851060224
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.416601, rank: 6, write(sync,parallel): 0.9993560314178467
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212500480, before: 1633570816, after: 1846071296
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 214368256, before: 1633574912, after: 1847943168
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.4318566, rank: 5, write(sync,parallel): 1.011610507965088
+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: 214933504, before: 1648107520, after: 1863041024
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212475904, before: 1614884864, after: 1827360768
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.4699008, rank: 3, write(sync,parallel): 0.9958508014678955
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 212373504, before: 1617059840, after: 1829433344
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.481593, rank: 7, write(sync,parallel): 1.064600944519043
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.499954, rank: 2, write(sync,parallel): 1.0260334014892578
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 214335488, before: 1617063936, after: 1831399424
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543458.5581553, rank: 1, write(sync,parallel): 1.0821151733398438
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, takes 6.389617919921875e-05 to finish D2H 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, takes 0.04180765151977539 to schedule async ckpt 
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 1, joining self.process
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 2, joining self.process
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.42s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.42s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 3, 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: 7, joining self.process
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.42s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.47s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.47s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 6, joining self.process
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 1.47s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 4, joining self.process
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 started
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:rank: 0, joining self.process
+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: 25505792, before: 1899667456, after: 1925173248
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 818884608, before: 1620201472, after: 2439086080
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1 consumed: 816308224, before: 1620213760, after: 2436521984
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543461.565737, rank: 4, write(sync,parallel): 3.2021517753601074
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 3.29s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:0 consumed: 1615142912, before: 1899667456, after: 3514810368
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:FileSystemWriterAsync: collected worker results successfully
+DEBUG:megatron.core.dist_checkpointing.strategies.filesystem_async:1750543464.5586467, rank: 0, write(sync,parallel): 5.696138143539429
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:TemporalAsyncCaller: Async process join finished after 5.78s from forking
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6032062, 1, gather: 5.740522146224976
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6032386, 3, gather: 5.740435838699341
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6033263, 6, gather: 5.7400062084198
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.603318, 5, gather: 5.740565776824951
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6032891, 2, gather: 5.741005897521973
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6035469, 7, gather: 5.740869045257568
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6084177, 0, gather: 0.008083581924438477
+DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0105s
+DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750543464.6084206, 4, gather: 2.9939053058624268
+Running ctx_length=131072, TP_SIZE=4, CP_SIZE=2, BATCH_SIZE=1
+Cleaning up checkpoint directory: gpt-checkpoint
+--------------------------------
+CTX_LENGTH: 131072
+TP_SIZE: 4
+CP_SIZE: 2
+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: 8, data-parallel size: 1, context-parallel size: 2, 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 ........................... 2
+  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 .............................. 131072
+  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 ..................... 8
+  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 ......................... 131072
+  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 ...................................... 131072
+  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 ...................................... 8
+  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
+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
+ > 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
+> 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: 0.046 seconds
+WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
+> compiling and loading fused kernels ...
+>>> done with compiling and loading fused kernels. Compilation time: 2.242 seconds
+time to initialize megatron (seconds): 7.855
+[after megatron is initialized] datetime: 2025-06-21 22:06:34 
+building GPT model ...
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (3, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (1, 0): 676924416
+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 (676924416 elements, 676924416 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.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.0.self_attention.linear_proj.weight
+	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_weight
+	module.decoder.layers.1.mlp.linear_fc2.bias
+	module.decoder.layers.1.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.0.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_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.embedding.position_embeddings.weight
+	module.decoder.layers.1.mlp.linear_fc2.weight
+	module.decoder.layers.1.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=<megatron.core.timers.Timers object at 0x14d42677eb10>, config_logger_dir='')
+INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (2, 0): 676924416
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 676924416
+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 ................................: (2.95, 3.93)
+[after model, optimizer, and learning rate scheduler are built] datetime: 2025-06-21 22:06:42 
+> 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=131072, 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=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x14d42671f050>, 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: 0.004888 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 520
+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: 0.001661 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 520
+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: 0.001424 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 520
+INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1
+> finished creating GPT datasets ...
+[after dataloaders are built] datetime: 2025-06-21 22:06:42 
+done with setup ...
+(min, max) time across ranks (ms):
+    model-and-optimizer-setup ......................: (8039.58, 8039.75)
+    train/valid/test-data-iterators-setup ..........: (21.56, 114.70)
+training ...
+Setting rerun_state_machine.current_iteration to 0...
+[before the start of training step] datetime: 2025-06-21 22:06:42 
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])
+batch tensor: tokens torch.Size([1, 131072])
+batch tensor: labels torch.Size([1, 131072])
+batch tensor: loss_mask torch.Size([1, 131072])
+batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
+batch tensor: position_ids torch.Size([1, 131072])
+batch tensor after cp: tokens torch.Size([1, 65536])
+batch tensor after cp: labels torch.Size([1, 65536])
+batch tensor after cp: loss_mask torch.Size([1, 65536])
+batch tensor after cp: attention_mask torch.Size([1, 1, 65536, 131072])
+batch tensor after cp: position_ids torch.Size([1, 65536])

attnserver.run_attnserver.slurm.sh.343226.err.log

@@ -1297,3 +1297,74 @@ W0621 21:55:51.007000 1996275 site-packages/torch/distributed/run.py:766] ******
   warnings.warn(
 /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
   warnings.warn(
+[rank2]:[W621 22:04:59.407544011 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank3]:[W621 22:04:59.608853506 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank7]:[W621 22:04:59.678669926 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank1]:[W621 22:04:59.700049367 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank6]:[W621 22:04:59.938906957 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank5]:[W621 22:05:00.259868129 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank0]:[W621 22:05:00.324680781 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank4]:[W621 22:05:00.380660785 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=81920
++ PROF_CTX_LENGTH=81920
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp4.cp2.bs2.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp4.cp2.bs2.json' ']'
++ echo 'Running ctx_length=81920, TP_SIZE=4, CP_SIZE=2, BATCH_SIZE=2'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343226 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-896:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 4 --context-parallel-size 2 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:05:16.304000 2000339 site-packages/torch/distributed/run.py:766] 
+W0621 22:05:16.304000 2000339 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:05:16.304000 2000339 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:05:16.304000 2000339 site-packages/torch/distributed/run.py:766] *****************************************
+[rank7]:[W621 22:05:38.315413964 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:05:38.315420577 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank1]:[W621 22:05:38.339507253 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:05:38.339635072 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:05:38.340067732 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:05:38.341018027 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:05:38.344861448 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:05:38.496511209 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(

attnserver.run_attnserver.slurm.sh.343237.err.log

@@ -2118,3 +2118,355 @@ W0621 21:59:08.581000 1124226 site-packages/torch/distributed/run.py:766] ******
 [rank13]:[W621 21:59:34.544805119 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
 [rank8]:[W621 21:59:34.671912210 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
 [rank0]:[W621 21:59:34.026347727 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
+[rank0]:     common_strategy.save_common(state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
+[rank0]:     torch.save(common_state_dict, path)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
+[rank0]:     with _open_zipfile_writer(f) as opened_zipfile:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
+[rank0]:     return container(name_or_buffer)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
+[rank0]:     torch._C.PyTorchFileWriter(
+[rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.
+[rank0]:[W621 22:04:21.659783891 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:04:31.929000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849652 closing signal SIGTERM
+W0621 22:04:31.932000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849653 closing signal SIGTERM
+W0621 22:04:31.934000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849654 closing signal SIGTERM
+W0621 22:04:31.937000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849655 closing signal SIGTERM
+W0621 22:04:31.940000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849656 closing signal SIGTERM
+W0621 22:04:31.943000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849657 closing signal SIGTERM
+W0621 22:04:31.960000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 849658 closing signal SIGTERM
+E0621 22:04:35.819000 849579 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 849651) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:04:31
+  host      : fs-mbz-gpu-274
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 849651)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
++ set +x
+W0621 22:04:36.170000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124296 closing signal SIGTERM
+W0621 22:04:36.171000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124297 closing signal SIGTERM
+W0621 22:04:36.176000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124298 closing signal SIGTERM
+W0621 22:04:36.179000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124299 closing signal SIGTERM
+W0621 22:04:36.182000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124300 closing signal SIGTERM
+W0621 22:04:36.200000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124301 closing signal SIGTERM
+W0621 22:04:36.204000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124302 closing signal SIGTERM
+W0621 22:04:36.213000 1124226 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1124303 closing signal SIGTERM
+[W621 22:04:39.022871617 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-476]:33082, remote=[fs-mbz-gpu-274]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14b0815785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14b06a45aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14b06a45c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14b06a45db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14b06a457ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14b06a457ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14b06a458f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14b07978b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14b078efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #17: <unknown function> + 0x94ac3 (0x14b082640ac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #18: <unknown function> + 0x126850 (0x14b0826d2850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:39.114000 1124226 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-476_1124226_0' has failed to send a keep-alive heartbeat to the rendezvous '343237' due to an error of type RendezvousConnectionError.
+[W621 22:04:39.362549495 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-476]:33082, remote=[fs-mbz-gpu-274]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14b0815785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14b06a45aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14b06a45c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14b06a45db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14b06a457ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14b06a457ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14b06a458f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14b07978b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14b078efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #26: <unknown function> + 0x29d90 (0x14b0825d5d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #27: __libc_start_main + 0x80 (0x14b0825d5e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:39.457000 1124226 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-476_1124226_0' has failed to shutdown the rendezvous '343237' due to an error of type RendezvousConnectionError.
+[W621 22:04:39.377055685 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-476]:33082, remote=[fs-mbz-gpu-274]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14b0815785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14b06a45aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14b06a45c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14b06a45db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14b06a457ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14b06a457ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14b06a458f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14b07978b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14b078efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #26: <unknown function> + 0x29d90 (0x14b0825d5d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #27: __libc_start_main + 0x80 (0x14b0825d5e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:39.469000 1124226 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-476_1124226_0' has failed to shutdown the rendezvous '343237' due to an error of type RendezvousConnectionError.
+Traceback (most recent call last):
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 117, in _call_store
+    return getattr(self._store, store_op)(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+torch.distributed.DistNetworkError: failed to recv, got 0 bytes
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
+    result = agent.run()
+             ^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
+    result = f(*args, **kwargs)
+             ^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
+    result = self._invoke_run(role)
+             ^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 906, in _invoke_run
+    num_nodes_waiting = rdzv_handler.num_nodes_waiting()
+                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 1263, in num_nodes_waiting
+    self._state_holder.sync()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 437, in sync
+    get_response = self._backend.get_state()
+                   ^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 75, in get_state
+    base64_state: bytes = self._call_store("get", self._key)
+                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 119, in _call_store
+    raise RendezvousConnectionError(
+torch.distributed.elastic.rendezvous.api.RendezvousConnectionError: The connection to the C10d store has failed. See inner exception for details.
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=81920
++ PROF_CTX_LENGTH=81920
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp8.bs1.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp8.bs1.json' ']'
++ echo 'Running ctx_length=81920, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=1'
++ srun bash ./attnserver.sh
++ which python3
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343237 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-274:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343237 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-274:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:05:02.807000 853291 site-packages/torch/distributed/run.py:766] 
+W0621 22:05:02.807000 853291 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:05:02.807000 853291 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:05:02.807000 853291 site-packages/torch/distributed/run.py:766] *****************************************
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:05:02.863000 1127779 site-packages/torch/distributed/run.py:766] 
+W0621 22:05:02.863000 1127779 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:05:02.863000 1127779 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:05:02.863000 1127779 site-packages/torch/distributed/run.py:766] *****************************************
+[rank9]:[W621 22:05:26.047374057 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank1]:[W621 22:05:26.386654497 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:05:26.387130751 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:05:26.387149164 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:05:26.387344347 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:05:26.387535395 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:05:26.388452688 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:05:26.388552862 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank11]:[W621 22:05:26.060819898 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank10]:[W621 22:05:26.060864338 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank15]:[W621 22:05:26.061040541 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank12]:[W621 22:05:26.061040709 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank14]:[W621 22:05:26.061049399 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank13]:[W621 22:05:26.061080704 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:05:26.543826674 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank8]:[W621 22:05:26.317480533 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(

attnserver.run_attnserver.slurm.sh.343238.err.log

@@ -6531,3 +6531,616 @@ W0621 21:59:31.558000 3515598 site-packages/torch/distributed/elastic/multiproce
 W0621 21:59:31.561000 3515598 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3515675 closing signal SIGTERM
 W0621 21:59:31.584000 3515598 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3515676 closing signal SIGTERM
 W0621 21:59:31.588000 3515598 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3515677 closing signal SIGTERM
+E0621 21:59:38.566000 3515598 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 3515670) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_21:59:31
+  host      : fs-mbz-gpu-518
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 3515670)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+W0621 21:59:38.797000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747805 closing signal SIGTERM
+W0621 21:59:38.800000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747806 closing signal SIGTERM
+W0621 21:59:38.803000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747807 closing signal SIGTERM
+W0621 21:59:38.805000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747808 closing signal SIGTERM
+W0621 21:59:38.808000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747809 closing signal SIGTERM
+W0621 21:59:38.812000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747810 closing signal SIGTERM
+W0621 21:59:38.815000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747811 closing signal SIGTERM
+W0621 21:59:38.817000 2747734 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2747812 closing signal SIGTERM
++ set +x
+[W621 21:59:39.196576404 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:59960, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14fb1c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14fb0585aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14fb0585c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14fb0585db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14fb05857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14fb05857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14fb05858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14fb14b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14fb142fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #17: <unknown function> + 0x94ac3 (0x14fb1d8abac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #18: <unknown function> + 0x126850 (0x14fb1d93d850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 21:59:39.729000 2747734 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2747734_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 21:59:44.206199521 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:59960, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14fb1c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14fb0585aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14fb0585c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14fb0585db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14fb05857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14fb05857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14fb05858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14fb14b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14fb142fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #17: <unknown function> + 0x94ac3 (0x14fb1d8abac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #18: <unknown function> + 0x126850 (0x14fb1d93d850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 21:59:44.737000 2747734 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2747734_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 21:59:47.893825051 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:59960, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14fb1c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14fb0585aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14fb0585c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14fb0585db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14fb05857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14fb05857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14fb05858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14fb14b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14fb142fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #26: <unknown function> + 0x29d90 (0x14fb1d840d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #27: __libc_start_main + 0x80 (0x14fb1d840e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 21:59:47.430000 2747734 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-546_2747734_0' has failed to shutdown the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 21:59:47.908207435 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:59960, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14fb1c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14fb0585aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14fb0585c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14fb0585db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x14fb05857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x14fb05857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x14fb05858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #7: <unknown function> + 0xc0f526 (0x14fb14b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #8: <unknown function> + 0x37f17d (0x14fb142fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #26: <unknown function> + 0x29d90 (0x14fb1d840d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #27: __libc_start_main + 0x80 (0x14fb1d840e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 21:59:47.441000 2747734 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-546_2747734_0' has failed to shutdown the rendezvous '343238' due to an error of type RendezvousConnectionError.
+Traceback (most recent call last):
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 117, in _call_store
+    return getattr(self._store, store_op)(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+torch.distributed.DistNetworkError: failed to recv, got 0 bytes
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
+    result = agent.run()
+             ^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
+    result = f(*args, **kwargs)
+             ^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
+    result = self._invoke_run(role)
+             ^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 906, in _invoke_run
+    num_nodes_waiting = rdzv_handler.num_nodes_waiting()
+                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 1263, in num_nodes_waiting
+    self._state_holder.sync()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 437, in sync
+    get_response = self._backend.get_state()
+                   ^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 75, in get_state
+    base64_state: bytes = self._call_store("get", self._key)
+                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 119, in _call_store
+    raise RendezvousConnectionError(
+torch.distributed.elastic.rendezvous.api.RendezvousConnectionError: The connection to the C10d store has failed. See inner exception for details.
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=49152
++ PROF_CTX_LENGTH=49152
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs2.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs2.json' ']'
++ echo 'Running ctx_length=49152, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=2'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343238 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-518:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343238 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-518:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 21:59:50.569000 2750889 site-packages/torch/distributed/run.py:766] 
+W0621 21:59:50.569000 2750889 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 21:59:50.569000 2750889 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 21:59:50.569000 2750889 site-packages/torch/distributed/run.py:766] *****************************************
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 21:59:50.576000 3518819 site-packages/torch/distributed/run.py:766] 
+W0621 21:59:50.576000 3518819 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 21:59:50.576000 3518819 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 21:59:50.576000 3518819 site-packages/torch/distributed/run.py:766] *****************************************
+[rank1]:[W621 22:00:12.751080946 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:00:12.751255399 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:00:12.752504114 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank13]:[W621 22:00:12.401091126 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank9]:[W621 22:00:12.401128675 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank15]:[W621 22:00:12.402117847 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:00:12.754737124 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:00:12.755589107 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:00:12.756058804 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:00:12.758028576 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank10]:[W621 22:00:12.418635679 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank11]:[W621 22:00:12.418699043 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank12]:[W621 22:00:12.418779495 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank14]:[W621 22:00:12.420380263 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank8]:[W621 22:00:13.525486738 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:00:13.927684252 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
+[rank0]:     common_strategy.save_common(state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
+[rank0]:     torch.save(common_state_dict, path)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
+[rank0]:     with _open_zipfile_writer(f) as opened_zipfile:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
+[rank0]:     return container(name_or_buffer)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
+[rank0]:     torch._C.PyTorchFileWriter(
+[rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.
+[rank0]:[W621 22:04:04.242273949 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:04:14.016000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518893 closing signal SIGTERM
+W0621 22:04:14.019000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518894 closing signal SIGTERM
+W0621 22:04:14.023000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518895 closing signal SIGTERM
+W0621 22:04:14.026000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518896 closing signal SIGTERM
+W0621 22:04:14.055000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518897 closing signal SIGTERM
+W0621 22:04:14.071000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518898 closing signal SIGTERM
+W0621 22:04:14.077000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3518899 closing signal SIGTERM
+E0621 22:04:26.395000 3518819 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 3518892) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:04:14
+  host      : fs-mbz-gpu-518
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 3518892)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+[W621 22:04:26.058717060 TCPStore.cpp:115] [c10d] recvVector failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): failed to recv, got 0 bytes
+Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa0d0 (0x14f77545c0d0 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5baa81d (0x14f77545c81d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: <unknown function> + 0x5bab4a9 (0x14f77545d4a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x1fb (0x14f7754574cb in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #6: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #7: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #16: <unknown function> + 0x94ac3 (0x14f78d661ac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #17: <unknown function> + 0x126850 (0x14f78d6f3850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:26.588000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2750889_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 22:04:26.155624461 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #24: <unknown function> + 0x29d90 (0x14f78d5f6d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #25: __libc_start_main + 0x80 (0x14f78d5f6e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
++ set +x
+W0621 22:04:26.693000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750960 closing signal SIGTERM
+W0621 22:04:26.695000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750961 closing signal SIGTERM
+W0621 22:04:26.698000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750962 closing signal SIGTERM
+W0621 22:04:26.701000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750963 closing signal SIGTERM
+W0621 22:04:26.703000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750964 closing signal SIGTERM
+W0621 22:04:26.716000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750965 closing signal SIGTERM
+W0621 22:04:26.733000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750966 closing signal SIGTERM
+W0621 22:04:26.757000 2750889 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2750967 closing signal SIGTERM
+[W621 22:04:31.065120559 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #15: <unknown function> + 0x94ac3 (0x14f78d661ac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #16: <unknown function> + 0x126850 (0x14f78d6f3850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:31.595000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2750889_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 22:04:36.071948889 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #15: <unknown function> + 0x94ac3 (0x14f78d661ac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #16: <unknown function> + 0x126850 (0x14f78d6f3850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:36.603000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2750889_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 22:04:41.079968518 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #15: <unknown function> + 0x94ac3 (0x14f78d661ac3 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #16: <unknown function> + 0x126850 (0x14f78d6f3850 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:41.608000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-546_2750889_0' has failed to send a keep-alive heartbeat to the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 22:04:42.644396618 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #24: <unknown function> + 0x29d90 (0x14f78d5f6d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #25: __libc_start_main + 0x80 (0x14f78d5f6e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:42.182000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-546_2750889_0' has failed to shutdown the rendezvous '343238' due to an error of type RendezvousConnectionError.
+[W621 22:04:42.660351204 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-546]:33082, remote=[fs-mbz-gpu-518]:29500): Broken pipe
+Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14f78c5785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x5ba8afe (0x14f77545aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #2: <unknown function> + 0x5baa358 (0x14f77545c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #3: <unknown function> + 0x5babb3e (0x14f77545db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&, std::vector<unsigned char, std::allocator<unsigned char> > const&) + 0x299 (0x14f775457569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
+frame #5: <unknown function> + 0xc0f919 (0x14f78478b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+frame #6: <unknown function> + 0x37f17d (0x14f783efb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
+<omitting python frames>
+frame #24: <unknown function> + 0x29d90 (0x14f78d5f6d90 in /lib/x86_64-linux-gnu/libc.so.6)
+frame #25: __libc_start_main + 0x80 (0x14f78d5f6e40 in /lib/x86_64-linux-gnu/libc.so.6)
+
+W0621 22:04:42.193000 2750889 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-546_2750889_0' has failed to shutdown the rendezvous '343238' due to an error of type RendezvousConnectionError.
+Traceback (most recent call last):
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 117, in _call_store
+    return getattr(self._store, store_op)(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+torch.distributed.DistNetworkError: Broken pipe
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
+    result = agent.run()
+             ^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
+    result = f(*args, **kwargs)
+             ^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
+    result = self._invoke_run(role)
+             ^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 906, in _invoke_run
+    num_nodes_waiting = rdzv_handler.num_nodes_waiting()
+                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 1263, in num_nodes_waiting
+    self._state_holder.sync()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 423, in sync
+    set_response = self._backend.set_state(state_bits, self._token)
+                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 100, in set_state
+    base64_state: bytes = self._call_store(
+                          ^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 119, in _call_store
+    raise RendezvousConnectionError(
+torch.distributed.elastic.rendezvous.api.RendezvousConnectionError: The connection to the C10d store has failed. See inner exception for details.
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=65536
++ PROF_CTX_LENGTH=65536
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp8.bs2.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp8.bs2.json' ']'
++ echo 'Running ctx_length=65536, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=2'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343238 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-518:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343238 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-518:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:04:45.279000 3522238 site-packages/torch/distributed/run.py:766] 
+W0621 22:04:45.279000 3522238 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:04:45.279000 3522238 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:04:45.279000 3522238 site-packages/torch/distributed/run.py:766] *****************************************
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:04:45.598000 2754239 site-packages/torch/distributed/run.py:766] 
+W0621 22:04:45.598000 2754239 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:04:45.598000 2754239 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:04:45.598000 2754239 site-packages/torch/distributed/run.py:766] *****************************************
+[rank1]:[W621 22:05:08.863273627 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:05:08.864178359 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank11]:[W621 22:05:08.511902344 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank9]:[W621 22:05:08.511901889 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:05:08.865951320 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:05:08.867427894 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank13]:[W621 22:05:08.518094352 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:05:08.874100135 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:05:08.874376739 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank15]:[W621 22:05:08.522487048 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:05:08.877914995 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank12]:[W621 22:05:08.528926215 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank14]:[W621 22:05:08.529104503 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank10]:[W621 22:05:08.529138894 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank8]:[W621 22:05:08.616025184 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:05:08.001813088 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(

attnserver.run_attnserver.slurm.sh.343239.err.log

@@ -1093,3 +1093,96 @@ Root Cause (first observed failure):
   traceback : Signal 6 (SIGABRT) received by PID 2138453
 ========================================================
 + set +x
+[rank14]:[F621 22:07:05.268307799 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 14] [PG ID 0 PG GUID 0(default_pg) Rank 14] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank10]:[F621 22:07:05.268564615 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 10] [PG ID 0 PG GUID 0(default_pg) Rank 10] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank12]:[F621 22:07:06.776354590 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 12] [PG ID 0 PG GUID 0(default_pg) Rank 12] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank13]:[F621 22:07:06.776790180 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 13] [PG ID 0 PG GUID 0(default_pg) Rank 13] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank9]:[F621 22:07:06.777117563 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 9] [PG ID 0 PG GUID 0(default_pg) Rank 9] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank11]:[F621 22:07:06.777319662 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 11] [PG ID 0 PG GUID 0(default_pg) Rank 11] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+[rank15]:[F621 22:07:06.778624888 ProcessGroupNCCL.cpp:1554] [PG ID 0 PG GUID 0(default_pg) Rank 15] [PG ID 0 PG GUID 0(default_pg) Rank 15] Terminating the process after attempting to dump debug info, due to collective timeout or exception.
+W0621 22:07:06.349000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792698 closing signal SIGTERM
+W0621 22:07:06.352000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792699 closing signal SIGTERM
+W0621 22:07:06.353000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792701 closing signal SIGTERM
+W0621 22:07:06.357000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792702 closing signal SIGTERM
+W0621 22:07:06.359000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792703 closing signal SIGTERM
+W0621 22:07:06.360000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 792705 closing signal SIGTERM
+E0621 22:07:07.268000 792627 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: -6) local_rank: 2 (pid: 792700) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+=======================================================
+./pretrain_gpt_profile.py FAILED
+-------------------------------------------------------
+Failures:
+[1]:
+  time      : 2025-06-21_22:07:06
+  host      : fs-mbz-gpu-188
+  rank      : 14 (local_rank: 6)
+  exitcode  : -6 (pid: 792704)
+  error_file: <N/A>
+  traceback : Signal 6 (SIGABRT) received by PID 792704
+-------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:07:06
+  host      : fs-mbz-gpu-188
+  rank      : 10 (local_rank: 2)
+  exitcode  : -6 (pid: 792700)
+  error_file: <N/A>
+  traceback : Signal 6 (SIGABRT) received by PID 792700
+=======================================================
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=12288
++ PROF_CTX_LENGTH=12288
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L12288*tp2.cp8.bs4.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L12288*tp2.cp8.bs4.json' ']'
++ echo 'Running ctx_length=12288, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 12288 --max-position-embeddings 12288 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 12288 --max-position-embeddings 12288 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:07:10.219000 796503 site-packages/torch/distributed/run.py:766] 
+W0621 22:07:10.219000 796503 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:07:10.219000 796503 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:07:10.219000 796503 site-packages/torch/distributed/run.py:766] *****************************************
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:07:10.319000 2142274 site-packages/torch/distributed/run.py:766] 
+W0621 22:07:10.319000 2142274 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:07:10.319000 2142274 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:07:10.319000 2142274 site-packages/torch/distributed/run.py:766] *****************************************

attnserver.run_attnserver.slurm.sh.343239.out.log

@@ -9973,3 +9973,22 @@ Params for bucket 1 (313079808 elements, 313079808 padded size):
 	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=<megatron.core.timers.Timers object at 0x151e86832a80>, config_logger_dir='')
 INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
+Running ctx_length=12288, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4
+Cleaning up checkpoint directory: gpt-checkpoint
+--------------------------------
+CTX_LENGTH: 12288
+TP_SIZE: 2
+CP_SIZE: 8
+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: 2
+CP_SIZE: 8
+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

attnserver.run_attnserver.slurm.sh.343243.err.log

@@ -2257,3 +2257,202 @@ W0621 21:58:29.557000 1699565 site-packages/torch/distributed/run.py:766] ******
   warnings.warn(
 /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
   warnings.warn(
+[rank1]:[W621 22:01:03.195093101 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank3]:[W621 22:01:04.317047889 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank5]:[W621 22:01:04.319372678 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank0]:[W621 22:01:04.825261599 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank4]:[W621 22:01:04.855555890 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank6]:[W621 22:01:04.875952870 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank2]:[W621 22:01:04.896135105 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank7]:[W621 22:01:04.933639941 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=65536
++ PROF_CTX_LENGTH=65536
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp4.bs1.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp4.bs1.json' ']'
++ echo 'Running ctx_length=65536, TP_SIZE=2, CP_SIZE=4, BATCH_SIZE=1'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343243 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-296:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 4 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:01:27.928000 1703273 site-packages/torch/distributed/run.py:766] 
+W0621 22:01:27.928000 1703273 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:01:27.928000 1703273 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:01:27.928000 1703273 site-packages/torch/distributed/run.py:766] *****************************************
+[rank1]:[W621 22:01:49.427826166 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:01:49.427824730 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:01:49.427862574 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:01:49.428141012 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:01:49.433752632 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:01:49.433981739 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:01:49.436614926 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:01:49.574653227 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
+[rank0]:     common_strategy.save_common(state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
+[rank0]:     torch.save(common_state_dict, path)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
+[rank0]:     with _open_zipfile_writer(f) as opened_zipfile:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
+[rank0]:     return container(name_or_buffer)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
+[rank0]:     torch._C.PyTorchFileWriter(
+[rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.
+[rank0]:[W621 22:05:13.224670980 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:05:22.565000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703344 closing signal SIGTERM
+W0621 22:05:22.567000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703345 closing signal SIGTERM
+W0621 22:05:22.571000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703346 closing signal SIGTERM
+W0621 22:05:22.574000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703347 closing signal SIGTERM
+W0621 22:05:22.577000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703348 closing signal SIGTERM
+W0621 22:05:22.607000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703349 closing signal SIGTERM
+W0621 22:05:22.630000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1703350 closing signal SIGTERM
+E0621 22:05:25.694000 1703273 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 1703343) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:05:22
+  host      : fs-mbz-gpu-296
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 1703343)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=81920
++ PROF_CTX_LENGTH=81920
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp4.bs1.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp4.bs1.json' ']'
++ echo 'Running ctx_length=81920, TP_SIZE=2, CP_SIZE=4, BATCH_SIZE=1'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343243 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-296:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 4 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:05:36.831000 1706637 site-packages/torch/distributed/run.py:766] 
+W0621 22:05:36.831000 1706637 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:05:36.831000 1706637 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:05:36.831000 1706637 site-packages/torch/distributed/run.py:766] *****************************************
+[rank7]:[W621 22:06:00.848046576 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:06:00.848057369 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:06:00.848088736 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank1]:[W621 22:06:00.848088743 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:06:00.858281884 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:06:00.864451804 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:06:00.871896915 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:06:00.145457891 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(

attnserver.run_attnserver.slurm.sh.343244.err.log

@@ -3506,3 +3506,433 @@ W0621 21:58:29.857000 455826 site-packages/torch/distributed/run.py:766] *******
   warnings.warn(
 /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
   warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
+[rank0]:     common_strategy.save_common(state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
+[rank0]:     torch.save(common_state_dict, path)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
+[rank0]:     with _open_zipfile_writer(f) as opened_zipfile:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
+[rank0]:     return container(name_or_buffer)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
+[rank0]:     torch._C.PyTorchFileWriter(
+[rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.
+[rank0]:[W621 22:02:15.081133392 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:02:21.964000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455898 closing signal SIGTERM
+W0621 22:02:21.972000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455899 closing signal SIGTERM
+W0621 22:02:21.975000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455900 closing signal SIGTERM
+W0621 22:02:21.978000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455901 closing signal SIGTERM
+W0621 22:02:21.981000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455902 closing signal SIGTERM
+W0621 22:02:21.999000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455903 closing signal SIGTERM
+W0621 22:02:22.004000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 455904 closing signal SIGTERM
+E0621 22:02:26.735000 455826 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 455897) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:02:21
+  host      : fs-mbz-gpu-898
+  rank      : 0 (local_rank: 0)
+  exitcode  : 1 (pid: 455897)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=49152
++ PROF_CTX_LENGTH=49152
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp4.bs2.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp4.bs2.json' ']'
++ echo 'Running ctx_length=49152, TP_SIZE=2, CP_SIZE=4, BATCH_SIZE=2'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343244 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-898:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 4 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:02:32.110000 459115 site-packages/torch/distributed/run.py:766] 
+W0621 22:02:32.110000 459115 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:02:32.110000 459115 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:02:32.110000 459115 site-packages/torch/distributed/run.py:766] *****************************************
+[rank1]:[W621 22:02:54.594917000 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:02:54.594917126 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:02:54.595014737 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:02:54.601411024 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank2]:[W621 22:02:54.611145419 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:02:54.611238887 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:02:54.614530675 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:02:54.768257391 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+[rank6]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__6_0.distcp'
+
+[rank6]: The above exception was the direct cause of the following exception:
+
+[rank6]: Traceback (most recent call last):
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank6]:     pretrain(
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank6]:     save_checkpoint(
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank6]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank6]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank6]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank6]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank6]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank6]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank6]:     finalize_fn()
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank6]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank6]:     write_results = storage_writer.retrieve_write_results()
+[rank6]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank6]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank6]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__6_0.distcp'
+[rank7]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__7_0.distcp'
+
+[rank7]: The above exception was the direct cause of the following exception:
+
+[rank7]: Traceback (most recent call last):
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank7]:     pretrain(
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank7]:     save_checkpoint(
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank7]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank7]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank7]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank7]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank7]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank7]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank7]:     finalize_fn()
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank7]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank7]:     write_results = storage_writer.retrieve_write_results()
+[rank7]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank7]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank7]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__7_0.distcp'
+[rank1]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__1_0.distcp'
+
+[rank1]: The above exception was the direct cause of the following exception:
+
+[rank1]: Traceback (most recent call last):
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank1]:     pretrain(
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank1]:     save_checkpoint(
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank1]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank1]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank1]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank1]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank1]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank1]:     finalize_fn()
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank1]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank1]:     write_results = storage_writer.retrieve_write_results()
+[rank1]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank1]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank1]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__1_0.distcp'
+[rank5]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__5_0.distcp'
+
+[rank5]: The above exception was the direct cause of the following exception:
+
+[rank5]: Traceback (most recent call last):
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank5]:     pretrain(
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank5]:     save_checkpoint(
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank5]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank5]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank5]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank5]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank5]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank5]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank5]:     finalize_fn()
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank5]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank5]:     write_results = storage_writer.retrieve_write_results()
+[rank5]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank5]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank5]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__5_0.distcp'
+[rank3]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__3_0.distcp'
+
+[rank3]: The above exception was the direct cause of the following exception:
+
+[rank3]: Traceback (most recent call last):
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank3]:     pretrain(
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank3]:     save_checkpoint(
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank3]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank3]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank3]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank3]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank3]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank3]:     finalize_fn()
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank3]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank3]:     write_results = storage_writer.retrieve_write_results()
+[rank3]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank3]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank3]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__3_0.distcp'
+[rank2]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__2_0.distcp'
+
+[rank2]: The above exception was the direct cause of the following exception:
+
+[rank2]: Traceback (most recent call last):
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank2]:     pretrain(
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank2]:     save_checkpoint(
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank2]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank2]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank2]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank2]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank2]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank2]:     finalize_fn()
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank2]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank2]:     write_results = storage_writer.retrieve_write_results()
+[rank2]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank2]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank2]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__2_0.distcp'
+[rank4]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__4_0.distcp'
+
+[rank4]: The above exception was the direct cause of the following exception:
+
+[rank4]: Traceback (most recent call last):
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank4]:     pretrain(
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank4]:     save_checkpoint(
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank4]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank4]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank4]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank4]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank4]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank4]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank4]:     finalize_fn()
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank4]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank4]:     write_results = storage_writer.retrieve_write_results()
+[rank4]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank4]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank4]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__4_0.distcp'
+[rank0]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__0_0.distcp'
+
+[rank0]: The above exception was the direct cause of the following exception:
+
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 404, in save
+[rank0]:     sharded_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/fully_parallel.py", line 95, in save
+[rank0]:     return self.base_strategy.save(sharded_state_dict, checkpoint_dir)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/base.py", line 228, in save
+[rank0]:     async_calls.maybe_finalize_async_calls(blocking=True)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/async_utils.py", line 545, in maybe_finalize_async_calls
+[rank0]:     finalize_fn()
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/torch.py", line 800, in finalize_fn
+[rank0]:     save_state_dict_async_finalize(*save_state_dict_ret)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/state_dict_saver.py", line 229, in save_state_dict_async_finalize
+[rank0]:     write_results = storage_writer.retrieve_write_results()
+[rank0]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/filesystem_async.py", line 436, in retrieve_write_results
+[rank0]:     raise RuntimeError(f'Worker failure: {write_results_or_exc}') from write_results_or_exc
+[rank0]: RuntimeError: Worker failure: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/__0_0.distcp'
+[rank5]:[W621 22:06:47.885164319 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank1]:[W621 22:06:47.900529072 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank7]:[W621 22:06:48.379264793 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+[rank3]:[W621 22:06:48.540156179 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+W0621 22:06:50.322000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459186 closing signal SIGTERM
+W0621 22:06:50.331000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459187 closing signal SIGTERM
+W0621 22:06:50.332000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459188 closing signal SIGTERM
+W0621 22:06:50.334000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459189 closing signal SIGTERM
+W0621 22:06:50.335000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459190 closing signal SIGTERM
+W0621 22:06:50.339000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459192 closing signal SIGTERM
+W0621 22:06:50.374000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 459193 closing signal SIGTERM
+E0621 22:07:10.031000 459115 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 459191) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
+    main()
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
+    return arg(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
+    launch(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
+    run(args)
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
+    elastic_launch(
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+./pretrain_gpt_profile.py FAILED
+------------------------------------------------------------
+Failures:
+  <NO_OTHER_FAILURES>
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-06-21_22:06:50
+  host      : fs-mbz-gpu-898
+  rank      : 5 (local_rank: 5)
+  exitcode  : 1 (pid: 459191)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
++ set +x
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=65536
++ PROF_CTX_LENGTH=65536
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp4.bs2.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp4.bs2.json' ']'
++ echo 'Running ctx_length=65536, TP_SIZE=2, CP_SIZE=4, BATCH_SIZE=2'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343244 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-898:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 4 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:07:13.434000 462698 site-packages/torch/distributed/run.py:766] 
+W0621 22:07:13.434000 462698 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:07:13.434000 462698 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:07:13.434000 462698 site-packages/torch/distributed/run.py:766] *****************************************

attnserver.run_attnserver.slurm.sh.343261.err.log

@@ -0,0 +1,202 @@
++ source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate
+++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda
+++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh
++++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
++++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
++++ export _CE_M=
++++ _CE_M=
++++ export _CE_CONDA=
++++ _CE_CONDA=
++++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
++++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
++++ '[' -z x ']'
+++ conda activate
+++ local cmd=activate
+++ case "$cmd" in
+++ __conda_activate activate
+++ '[' -n '' ']'
+++ local ask_conda
++++ PS1=
++++ __conda_exe shell.posix activate
++++ '[' -n '' ']'
++++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate
+++ ask_conda='unset _CE_M
+unset _CE_CONDA
+PS1='\''(base) '\''
+export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
+export CONDA_SHLVL='\''1'\''
+export CONDA_PROMPT_MODIFIER='\''(base) '\''
+export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
+export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
+++ eval 'unset _CE_M
+unset _CE_CONDA
+PS1='\''(base) '\''
+export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
+export CONDA_SHLVL='\''1'\''
+export CONDA_PROMPT_MODIFIER='\''(base) '\''
+export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
+export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
++++ unset _CE_M
++++ unset _CE_CONDA
++++ PS1='(base) '
++++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
++++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
++++ export CONDA_SHLVL=1
++++ CONDA_SHLVL=1
++++ export 'CONDA_PROMPT_MODIFIER=(base) '
++++ CONDA_PROMPT_MODIFIER='(base) '
++++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
++++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
++++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
++++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
+++ __conda_hashr
+++ '[' -n '' ']'
+++ '[' -n '' ']'
+++ hash -r
++ conda activate junda-attnserver
++ local cmd=activate
++ case "$cmd" in
++ __conda_activate activate junda-attnserver
++ '[' -n '' ']'
++ local ask_conda
+++ PS1='(base) '
+++ __conda_exe shell.posix activate junda-attnserver
+++ '[' -n '' ']'
+++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver
++ ask_conda='unset _CE_M
+unset _CE_CONDA
+PS1='\''(junda-attnserver) '\''
+export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
+export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
+export CONDA_SHLVL='\''2'\''
+export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
+export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
+export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
+export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
+export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
++ eval 'unset _CE_M
+unset _CE_CONDA
+PS1='\''(junda-attnserver) '\''
+export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
+export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
+export CONDA_SHLVL='\''2'\''
+export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
+export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
+export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
+export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
+export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
+++ unset _CE_M
+++ unset _CE_CONDA
+++ PS1='(junda-attnserver) '
+++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
+++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
+++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
+++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
+++ export CONDA_SHLVL=2
+++ CONDA_SHLVL=2
+++ export CONDA_DEFAULT_ENV=junda-attnserver
+++ CONDA_DEFAULT_ENV=junda-attnserver
+++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) '
+++ CONDA_PROMPT_MODIFIER='(junda-attnserver) '
+++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
+++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
+++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
+++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
+++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
+++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
++ __conda_hashr
++ '[' -n '' ']'
++ '[' -n '' ']'
++ hash -r
++ export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
++ CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
++ mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
++ export PROF_TP_SIZE=1
++ PROF_TP_SIZE=1
++ export PROF_CP_SIZE=8
++ PROF_CP_SIZE=8
++ export PROF_BS=1
++ PROF_BS=1
++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
++ export PROF_CTX_LENGTH=1024
++ PROF_CTX_LENGTH=1024
++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp1.cp8.bs1.json'
++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp1.cp8.bs1.json' ']'
++ echo 'Running ctx_length=1024, TP_SIZE=1, CP_SIZE=8, BATCH_SIZE=1'
++ srun bash ./attnserver.sh
++ which python3
++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343261 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-830:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 1 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated
+and will be removed in future. Use torchrun.
+Note that --use-env is set by default in torchrun.
+If your script expects `--local-rank` argument to be set, please
+change it to read from `os.environ['LOCAL_RANK']` instead. See 
+https://pytorch.org/docs/stable/distributed.html#launch-utility for 
+further instructions
+
+  main()
+W0621 22:06:13.082000 2070539 site-packages/torch/distributed/run.py:766] 
+W0621 22:06:13.082000 2070539 site-packages/torch/distributed/run.py:766] *****************************************
+W0621 22:06:13.082000 2070539 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0621 22:06:13.082000 2070539 site-packages/torch/distributed/run.py:766] *****************************************
+[rank2]:[W621 22:06:35.957817612 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank5]:[W621 22:06:35.957825745 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank1]:[W621 22:06:35.957834527 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank3]:[W621 22:06:35.958544022 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank4]:[W621 22:06:35.960944235 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank6]:[W621 22:06:35.963661455 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank7]:[W621 22:06:35.963839061 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+[rank0]:[W621 22:06:35.101598532 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
+  warnings.warn(
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
+[rank0]:     pretrain(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
+[rank0]:     save_checkpoint(
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
+[rank0]:     async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
+[rank0]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
+[rank0]:     common_strategy.save_common(state_dict, checkpoint_dir)
+[rank0]:   File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
+[rank0]:     torch.save(common_state_dict, path)
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
+[rank0]:     with _open_zipfile_writer(f) as opened_zipfile:
+[rank0]:          ^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
+[rank0]:     return container(name_or_buffer)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
+[rank0]:     torch._C.PyTorchFileWriter(
+[rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.

attnserver.run_attnserver.slurm.sh.343261.out.log

@@ -0,0 +1,1507 @@
+Running ctx_length=1024, TP_SIZE=1, CP_SIZE=8, BATCH_SIZE=1
+Cleaning up checkpoint directory: gpt-checkpoint
+--------------------------------
+CTX_LENGTH: 1024
+TP_SIZE: 1
+CP_SIZE: 8
+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
+using world size: 8, data-parallel size: 1, context-parallel size: 8, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 1, 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 ........................... 8
+  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 ..................... 1
+  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 ..................... 8
+  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 ...................... 1
+  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 ...................................... 8
+  yaml_cfg ........................................ None
+-------------------- end of arguments ---------------------
+INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
+> building GPT2BPETokenizer tokenizer ...
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
+INFO:megatron.training.initialize:Setting logging level to 0
+WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED
+> initializing torch distributed ...
+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
+> initialized tensor model parallel with size 1
+> 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: 0.059 seconds
+> compiling and loading fused kernels ...
+>>> done with compiling and loading fused kernels. Compilation time: 2.515 seconds
+time to initialize megatron (seconds): 7.924
+[after megatron is initialized] datetime: 2025-06-21 22:06:42 
+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 (0, 0): 562663424
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+>>> embedding
+>>> decoder
+>>> output_layer
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 562663424
+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 (562663424 elements, 562663424 padded size):
+	module.decoder.final_layernorm.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.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.layers.0.self_attention.linear_proj.weight
+	module.embedding.word_embeddings.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.0.self_attention.linear_proj.bias
+	module.decoder.layers.1.mlp.linear_fc1.weight
+	module.decoder.layers.0.mlp.linear_fc1.weight
+	module.embedding.position_embeddings.weight
+	module.decoder.layers.1.mlp.linear_fc2.bias
+	module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
+	module.decoder.final_layernorm.bias
+	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_qkv.layer_norm_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=<megatron.core.timers.Timers object at 0x14972775eea0>, 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 ................................: (3.32, 3.38)
+[after model, optimizer, and learning rate scheduler are built] datetime: 2025-06-21 22:06:42 
+> 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=1024, 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=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x14972781f770>, 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: 0.005874 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 66592
+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: 0.003374 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 66562
+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: 0.003366 seconds
+INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 66686
+INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1
+> finished creating GPT datasets ...
+[after dataloaders are built] datetime: 2025-06-21 22:06:42 
+done with setup ...
+training ...(min, max) time across ranks (ms):
+    model-and-optimizer-setup ......................: (271.13, 289.44)
+    train/valid/test-data-iterators-setup ..........: (168.82, 202.81)
+
+Setting rerun_state_machine.current_iteration to 0...
+[before the start of training step] datetime: 2025-06-21 22:06:43 
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor:batch tensor: labels  torch.Size([1, 1024])tokens
+ batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_masktorch.Size([1, 1024]) torch.Size([1, 1, 1024, 1024])
+
+batch tensor: position_ids batch tensor:torch.Size([1, 1024])
+ labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 0
+Done exporting trace 0
+Number of parameters in transformer block in billions:  0.35
+ [2025-06-21 22:06:58] iteration        1/      10 | consumed samples:            1 | elapsed time per iteration (ms): 15326.1 | learning rate: 0.000000E+00 | global batch size:     1 | loss scale: 4294967296.0 | number of skipped iterations:   1 | number of nan iterations:   0 |
+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: 0.5584
+Theoretical memory footprints: weight and optimizer=9585.70 MB
+[Rank 2] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0
+[Rank 5] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7224.0 | max reserved: 7224.0
+[Rank 4] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0[Rank 0] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0[Rank 1] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0
+
+
+[Rank 3] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0[Rank 7] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7224.0 | max reserved: 7224.0
+
+[Rank 6] (after 1 iterations) memory (MB) | allocated: 6874.138671875 | max allocated: 6874.1396484375 | reserved: 7222.0 | max reserved: 7222.0
+batch tensor: tokens batch tensor: tokenstorch.Size([1, 1024]) 
+batch tensor: labels torch.Size([1, 1024])torch.Size([1, 1024])
+
+batch tensor: loss_maskbatch tensor:  torch.Size([1, 1024])labels
+ torch.Size([1, 1024])
+batch tensor: batch tensor:attention_mask  loss_masktorch.Size([1, 1, 1024, 1024]) 
+torch.Size([1, 1024])
+batch tensor: batch tensor:position_ids  attention_masktorch.Size([1, 1024]) 
+torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor:batch tensor: position_ids  torch.Size([1, 1024])tokens
+ torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])batch tensor:
+ batch tensor:tokens  labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+torch.Size([1, 1024])batch tensor:
+ attention_mask batch tensor:torch.Size([1, 1, 1024, 1024]) 
+labels batch tensor:torch.Size([1, 1024]) 
+position_ids batch tensor:torch.Size([1, 1024]) 
+loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: batch tensor after cp:loss_mask  tokenstorch.Size([1, 128]) 
+batch tensor after cp:torch.Size([1, 128]) 
+attention_mask batch tensor after cp: torch.Size([1, 1, 128, 1024])labels
+ batch tensor after cp:torch.Size([1, 128]) 
+position_idsbatch tensor after cp:  torch.Size([1, 128])loss_mask
+ torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp:batch tensor after cp:  tokenstokens  torch.Size([1, 128])
+batch tensor after cp:torch.Size([1, 128]) 
+labels batch tensor after cp:torch.Size([1, 128]) 
+labelsbatch tensor after cp:  torch.Size([1, 128])loss_mask
+ batch tensor after cp:torch.Size([1, 128]) 
+loss_maskbatch tensor after cp:  torch.Size([1, 128])attention_mask
+ batch tensor after cp:torch.Size([1, 1, 128, 1024]) 
+attention_mask batch tensor after cp: torch.Size([1, 1, 128, 1024])position_ids
+ batch tensor after cp:torch.Size([1, 128]) 
+position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 1
+Done exporting trace 1
+ [2025-06-21 22:06:58] iteration        2/      10 | consumed samples:            2 | elapsed time per iteration (ms): 75.7 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: batch tensor:attention_maskbatch tensor: batch tensor:  torch.Size([1, 1, 1024, 1024]) tokens
+tokensbatch tensor:tokens    position_idstorch.Size([1, 1024]) 
+torch.Size([1, 1024])torch.Size([1, 1024])batch tensor:torch.Size([1, 1024])
+
+batch tensor: 
+batch tensor: tokens labelsbatch tensor: labels   torch.Size([1, 1024])torch.Size([1, 1024])labels
+
+torch.Size([1, 1024])batch tensor: batch tensor: 
+loss_masktorch.Size([1, 1024])  batch tensor:loss_mask 
+torch.Size([1, 1024]) labels batch tensor:batch tensor: 
+torch.Size([1, 1024])torch.Size([1, 1024]) loss_mask
+
+batch tensor:tokens  batch tensor:batch tensor: torch.Size([1, 1024]) attention_maskloss_mask  torch.Size([1, 1024])
+ attention_masktorch.Size([1, 1, 1024, 1024])batch tensor:
+
+torch.Size([1, 1024])  batch tensor:attention_mask  
+batch tensor:batch tensor:batch tensor:torch.Size([1, 1, 1024, 1024]) 
+attention_mask batch tensor:tokens position_idstorch.Size([1, 1, 1024, 1024])  position_ids 
+batch tensor: batch tensor:labels  torch.Size([1, 1, 1024, 1024])torch.Size([1, 1024])
+ torch.Size([1, 1024])position_idstokens torch.Size([1, 1024])
+batch tensor:
+  torch.Size([1, 1024])torch.Size([1, 1024])batch tensor:
+
+ 
+torch.Size([1, 1024]) position_idslabelsbatch tensor: 
+  torch.Size([1, 1024])loss_masktorch.Size([1, 1024])batch tensor:
+
+  batch tensor:torch.Size([1, 1024])labels 
+ loss_maskbatch tensor:torch.Size([1, 1024])  
+attention_masktorch.Size([1, 1024])batch tensor:
+  loss_masktorch.Size([1, 1, 1024, 1024]) batch tensor:
+torch.Size([1, 1024]) batch tensor:
+attention_mask batch tensor: position_ids torch.Size([1, 1, 1024, 1024]) attention_mask
+batch tensor:torch.Size([1, 1024])  position_ids
+torch.Size([1, 1, 1024, 1024]) 
+torch.Size([1, 1024])batch tensor:
+ position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp:batch tensor after cp:batch tensor after cp:   batch tensor after cp:tokenstokenstokens   tokens torch.Size([1, 128]) torch.Size([1, 128])
+torch.Size([1, 128])
+torch.Size([1, 128])batch tensor after cp:
+
+batch tensor after cp:  batch tensor after cp:batch tensor after cp:labelslabels    labelstorch.Size([1, 128])labels
+torch.Size([1, 128]) batch tensor after cp: 
+torch.Size([1, 128]) batch tensor after cp:
+loss_masktorch.Size([1, 128])batch tensor after cp:  
+loss_mask torch.Size([1, 128])batch tensor after cp: loss_mask
+ torch.Size([1, 128]) batch tensor after cp:
+loss_masktorch.Size([1, 128])batch tensor after cp: batch tensor after cp: 
+torch.Size([1, 128])
+ batch tensor after cp:batch tensor after cp:attention_mask tokens attention_mask   torch.Size([1, 1, 128, 1024])batch tensor after cp:torch.Size([1, 128])attention_mask tokens
+ 
+ batch tensor after cp:torch.Size([1, 1, 128, 1024])batch tensor after cp:attention_mask  torch.Size([1, 1, 128, 1024])batch tensor after cp:
+ tokens
+torch.Size([1, 128])  batch tensor after cp:torch.Size([1, 1, 128, 1024])
+ batch tensor after cp:position_idslabels 
+batch tensor after cp: torch.Size([1, 128])  position_idsbatch tensor after cp: position_ids
+torch.Size([1, 128])torch.Size([1, 128])   
+labelstorch.Size([1, 128])batch tensor after cp:
+position_idstorch.Size([1, 128])batch tensor after cp:
+   
+ torch.Size([1, 128])loss_masklabelstorch.Size([1, 128])
+
+ batch tensor after cp: torch.Size([1, 128]) torch.Size([1, 128])
+loss_mask
+batch tensor after cp:  batch tensor after cp:torch.Size([1, 128]) attention_mask
+ loss_maskbatch tensor after cp:torch.Size([1, 1, 128, 1024])  
+torch.Size([1, 128])attention_mask
+batch tensor after cp: batch tensor after cp:  torch.Size([1, 1, 128, 1024])position_idsattention_mask
+  batch tensor after cp:torch.Size([1, 128])torch.Size([1, 1, 128, 1024]) 
+
+position_idsbatch tensor after cp:  torch.Size([1, 128])position_ids
+ torch.Size([1, 128])
+Start exporting trace 2
+Done exporting trace 2
+ [2025-06-21 22:06:58] iteration        3/      10 | consumed samples:            3 | elapsed time per iteration (ms): 48.6 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask batch tensor:torch.Size([1, 1, 1024, 1024])
+ batch tensor: tokensposition_ids  torch.Size([1, 1024])
+torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels batch tensor after cp:torch.Size([1, 1024])
+ tokensbatch tensor:  loss_mask torch.Size([1, 128])
+torch.Size([1, 1024])batch tensor after cp:batch tensor after cp:
+  batch tensor:tokens labels attention_mask torch.Size([1, 128]) torch.Size([1, 128])torch.Size([1, 1, 1024, 1024])
+
+
+batch tensor after cp:batch tensor after cp:batch tensor:batch tensor:   loss_mask  labelstorch.Size([1, 128]) 
+position_idstokens torch.Size([1, 128])
+batch tensor after cp: torch.Size([1, 1024])batch tensor after cp: 
+torch.Size([1, 1024]) attention_mask
+loss_mask  batch tensor:torch.Size([1, 1, 128, 1024])torch.Size([1, 128])
+ 
+batch tensor after cp:labelsbatch tensor after cp:   position_idstorch.Size([1, 1024]) attention_mask
+ torch.Size([1, 128])batch tensor:torch.Size([1, 1, 128, 1024])
+ 
+loss_mask batch tensor after cp: torch.Size([1, 1024])position_ids
+ torch.Size([1, 128])batch tensor:
+ attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 3
+Done exporting trace 3
+ [2025-06-21 22:06:58] iteration        4/      10 | consumed samples:            4 | elapsed time per iteration (ms): 45.5 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: batch tensor:loss_mask torch.Size([1, 1024]) 
+tokensbatch tensor:  attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: torch.Size([1, 1024])position_ids 
+torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens batch tensor:torch.Size([1, 1024])
+ tokensbatch tensor:  labels torch.Size([1, 1024])
+batch tensor: torch.Size([1, 1024])loss_mask 
+torch.Size([1, 1024])
+batch tensor:batch tensor:  labelsattention_mask  torch.Size([1, 1024])torch.Size([1, 1, 1024, 1024])
+
+batch tensor: batch tensor:loss_mask  position_idstorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp:batch tensor after cp:  loss_masktokens  torch.Size([1, 128])
+torch.Size([1, 128])batch tensor after cp:
+ attention_maskbatch tensor after cp:  torch.Size([1, 1, 128, 1024])labels
+ batch tensor after cp:torch.Size([1, 128]) 
+position_idsbatch tensor after cp: torch.Size([1, 128]) 
+loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor:batch tensor: batch tensor: tokens tokens  tokenstorch.Size([1, 1024]) 
+torch.Size([1, 1024])batch tensor:
+ torch.Size([1, 1024])labelsbatch tensor:
+  torch.Size([1, 1024])labels
+batch tensor: batch tensor:torch.Size([1, 1024])  
+labelsloss_maskbatch tensor:   torch.Size([1, 1024])loss_masktorch.Size([1, 1024])
+ 
+batch tensor:torch.Size([1, 1024])batch tensor: 
+ attention_maskbatch tensor:loss_mask  attention_mask  torch.Size([1, 1, 1024, 1024])torch.Size([1, 1024])
+torch.Size([1, 1, 1024, 1024])
+batch tensor:
+batch tensor:  attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids position_ids torch.Size([1, 1024])
+batch tensor:torch.Size([1, 1024]) 
+position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask batch tensor after cp:torch.Size([1, 128])batch tensor after cp: 
+tokens batch tensor after cp:tokens   torch.Size([1, 128])attention_masktorch.Size([1, 128])
+ 
+torch.Size([1, 1, 128, 1024])batch tensor after cp:batch tensor after cp: 
+ batch tensor after cp:labels labels position_ids torch.Size([1, 128]) 
+torch.Size([1, 128])torch.Size([1, 128])batch tensor after cp:
+
+ loss_maskbatch tensor after cp:  torch.Size([1, 128])loss_mask 
+torch.Size([1, 128])batch tensor after cp:
+ batch tensor after cp:attention_mask  attention_mask torch.Size([1, 1, 128, 1024])
+torch.Size([1, 1, 128, 1024])batch tensor after cp:
+ batch tensor after cp:position_ids  position_idstorch.Size([1, 128]) 
+torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 4
+Done exporting trace 4
+ [2025-06-21 22:06:58] iteration        5/      10 | consumed samples:            5 | elapsed time per iteration (ms): 44.8 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: batch tensor:position_ids torch.Size([1, 1024]) 
+tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: batch tensor after cp:tokens tokens  torch.Size([1, 128])
+torch.Size([1, 1024])batch tensor after cp:
+ labels batch tensor:torch.Size([1, 128]) 
+labelsbatch tensor after cp:  torch.Size([1, 1024])
+loss_maskbatch tensor:  torch.Size([1, 128])loss_mask
+ batch tensor after cp:torch.Size([1, 1024]) 
+attention_mask batch tensor:torch.Size([1, 1, 128, 1024]) 
+attention_maskbatch tensor after cp:  position_idstorch.Size([1, 1, 1024, 1024]) torch.Size([1, 128])
+
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor:batch tensor: tokens  tokens torch.Size([1, 1024])
+torch.Size([1, 1024])
+batch tensor: labelsbatch tensor:  labelstorch.Size([1, 1024]) 
+torch.Size([1, 1024])batch tensor:
+ batch tensor:loss_mask  loss_masktorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: batch tensor:attention_mask  attention_masktorch.Size([1, 1, 1024, 1024]) 
+torch.Size([1, 1, 1024, 1024])batch tensor:
+ position_idsbatch tensor:  position_idstorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor after cp: batch tensor:tokens  labels torch.Size([1, 128])torch.Size([1, 1024])
+
+batch tensor after cp:batch tensor:  labelsloss_mask  torch.Size([1, 128])torch.Size([1, 1024])
+
+batch tensor after cp:batch tensor:  loss_maskattention_mask  torch.Size([1, 128])torch.Size([1, 1, 1024, 1024])
+
+batch tensor after cp:batch tensor:  attention_maskposition_ids  torch.Size([1, 1, 128, 1024])torch.Size([1, 1024])
+
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens batch tensor after cp:torch.Size([1, 128]) 
+tokensbatch tensor after cp:  torch.Size([1, 128])labels
+ torch.Size([1, 128])batch tensor after cp:
+ batch tensor after cp:labels  loss_masktorch.Size([1, 128]) 
+torch.Size([1, 128])batch tensor after cp:
+ loss_maskbatch tensor after cp:  torch.Size([1, 128])attention_mask
+ batch tensor after cp:torch.Size([1, 1, 128, 1024]) 
+attention_mask batch tensor after cp:torch.Size([1, 1, 128, 1024]) 
+position_ids batch tensor after cp:torch.Size([1, 128]) 
+position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 5
+Done exporting trace 5
+ [2025-06-21 22:06:58] iteration        6/      10 | consumed samples:            6 | elapsed time per iteration (ms): 45.8 | 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:batch tensor:  tokenstokens  torch.Size([1, 1024])torch.Size([1, 1024])
+
+batch tensor: labelsbatch tensor:  torch.Size([1, 1024])labels
+batch tensor:  loss_masktorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: batch tensor:loss_mask  attention_masktorch.Size([1, 1024]) batch tensor:torch.Size([1, 1, 1024, 1024])
+
+ batch tensor:batch tensor: batch tensor:tokensposition_ids   attention_mask torch.Size([1, 1024]) 
+torch.Size([1, 1, 1024, 1024])tokens
+ torch.Size([1, 1024])batch tensor: 
+position_ids batch tensor:torch.Size([1, 1024])torch.Size([1, 1024]) 
+
+labels torch.Size([1, 1024])batch tensor:
+ batch tensor:labels  loss_masktorch.Size([1, 1024]) 
+torch.Size([1, 1024])batch tensor:
+ loss_maskbatch tensor:  torch.Size([1, 1024])attention_mask
+ torch.Size([1, 1, 1024, 1024])batch tensor:
+ attention_maskbatch tensor:  position_idstorch.Size([1, 1, 1024, 1024]) 
+torch.Size([1, 1024])batch tensor:
+ position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp:batch tensor after cp:  tokenslabels  batch tensor after cp:torch.Size([1, 128])torch.Size([1, 128]) 
+
+tokensbatch tensor after cp:batch tensor after cp:  loss_mask  labelstorch.Size([1, 128]) torch.Size([1, 128])
+torch.Size([1, 128])
+
+batch tensor after cp:batch tensor after cp:batch tensor after cp:   labelsattention_maskloss_mask batch tensor:  torch.Size([1, 128])
+torch.Size([1, 128])torch.Size([1, 1, 128, 1024])
+ batch tensor after cp:
+batch tensor after cp:tokens  batch tensor after cp:attention_mask loss_mask  position_ids  torch.Size([1, 1, 128, 1024])torch.Size([1, 128])torch.Size([1, 128])
+
+torch.Size([1, 1024])
+batch tensor after cp:batch tensor after cp:  
+attention_maskposition_ids  batch tensor:torch.Size([1, 1, 128, 1024])torch.Size([1, 128]) 
+
+labelsbatch tensor after cp:  torch.Size([1, 1024])position_ids
+ torch.Size([1, 128])batch tensor:
+ loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 6
+Done exporting trace 6
+ [2025-06-21 22:06:58] iteration        7/      10 | consumed samples:            7 | elapsed time per iteration (ms): 43.7 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor:batch tensor: labels  torch.Size([1, 1024])tokens
+ torch.Size([1, 1024])
+batch tensor:batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_maskbatch tensor: tokens torch.Size([1, 1024]) loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024]) torch.Size([1, 1024])
+batch tensor: attention_mask 
+torch.Size([1, 1, 1024, 1024])
+
+batch tensor:batch tensor:  batch tensor:position_idslabels   position_idstorch.Size([1, 1024])torch.Size([1, 1024]) 
+
+torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask batch tensor after cp:torch.Size([1, 1, 1024, 1024]) batch tensor after cp:tokens
+  batch tensor:batch tensor after cp: tokens torch.Size([1, 128])position_ids tokens
+ torch.Size([1, 128]) torch.Size([1, 128])batch tensor after cp:
+ torch.Size([1, 1024])batch tensor after cp:
+labels
+ batch tensor after cp: labels torch.Size([1, 128]) labels
+torch.Size([1, 128]) batch tensor after cp:
+torch.Size([1, 128]) 
+batch tensor after cp:batch tensor after cp:loss_mask   loss_maskloss_mask torch.Size([1, 128]) torch.Size([1, 128])
+torch.Size([1, 128])
+batch tensor after cp:
+batch tensor after cp: batch tensor after cp: attention_mask attention_mask attention_mask  torch.Size([1, 1, 128, 1024])torch.Size([1, 1, 128, 1024])torch.Size([1, 1, 128, 1024])
+
+
+batch tensor after cp:batch tensor after cp:  batch tensor after cp:position_ids position_idsposition_ids   batch tensor:batch tensor after cp:torch.Size([1, 128]) torch.Size([1, 128])tokenstorch.Size([1, 128])
+ 
+ 
+tokens torch.Size([1, 128])torch.Size([1, 1024])
+
+batch tensor after cp:batch tensor:  labelslabels  torch.Size([1, 1024])torch.Size([1, 128])
+
+batch tensor:batch tensor after cp:  loss_maskloss_mask  torch.Size([1, 1024])torch.Size([1, 128])
+
+batch tensor:batch tensor after cp:  attention_maskattention_mask  torch.Size([1, 1, 1024, 1024])
+torch.Size([1, 1, 128, 1024])
+batch tensor: batch tensor after cp: position_idsposition_ids  torch.Size([1, 1024])torch.Size([1, 128])
+
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 7
+Done exporting trace 7
+ [2025-06-21 22:06:58] iteration        8/      10 | consumed samples:            8 | elapsed time per iteration (ms): 43.3 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor:batch tensor: attention_mask  torch.Size([1, 1, 1024, 1024])tokens
+batch tensor:batch tensor:  position_ids  torch.Size([1, 1024])torch.Size([1, 1024])tokens
+
+ batch tensor: labels torch.Size([1, 1024])
+torch.Size([1, 1024])batch tensor: 
+loss_mask torch.Size([1, 1024])
+batch tensor:batch tensor:  labelsattention_mask  torch.Size([1, 1, 1024, 1024])torch.Size([1, 1024])
+
+batch tensor:batch tensor:  position_idsloss_mask  torch.Size([1, 1024])torch.Size([1, 1024])
+
+batch tensor:batch tensor: attention_mask  torch.Size([1, 1, 1024, 1024])tokens
+ batch tensor: position_ids torch.Size([1, 1024])torch.Size([1, 1024])
+
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor:batch tensor: loss_mask  torch.Size([1, 1024])tokens
+ batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+torch.Size([1, 1024])batch tensor:
+ position_ids batch tensor:torch.Size([1, 1024]) 
+labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])batch tensor after cp:
+ batch tensor after cp:tokens  attention_mask torch.Size([1, 128])torch.Size([1, 1, 128, 1024])
+
+batch tensor after cp:batch tensor after cp:  labelsposition_ids  torch.Size([1, 128])torch.Size([1, 128])
+
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp:batch tensor after cp:  tokensposition_ids  torch.Size([1, 128])torch.Size([1, 128])
+
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor:batch tensor:  tokenstokens  torch.Size([1, 1024])torch.Size([1, 1024])
+
+batch tensor: batch tensor:labels  labelstorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: batch tensor:loss_mask  loss_masktorch.Size([1, 1024]) 
+torch.Size([1, 1024])
+batch tensor: batch tensor:attention_mask  attention_mask torch.Size([1, 1, 1024, 1024])
+torch.Size([1, 1, 1024, 1024])
+batch tensor: batch tensor:position_ids  position_ids torch.Size([1, 1024])
+torch.Size([1, 1024])
+batch tensor after cp: tokens batch tensor after cp:torch.Size([1, 128]) 
+tokensbatch tensor after cp:  labelstorch.Size([1, 128]) 
+torch.Size([1, 128])batch tensor after cp:
+ batch tensor after cp:labels  loss_masktorch.Size([1, 128]) 
+torch.Size([1, 128])
+batch tensor after cp: batch tensor after cp:loss_mask  attention_masktorch.Size([1, 128]) 
+torch.Size([1, 1, 128, 1024])batch tensor after cp:
+ attention_maskbatch tensor after cp:  position_idstorch.Size([1, 1, 128, 1024]) 
+torch.Size([1, 128])batch tensor after cp:
+ position_ids torch.Size([1, 128])
+Start exporting trace 8
+Done exporting trace 8
+ [2025-06-21 22:06:58] iteration        9/      10 | consumed samples:            9 | elapsed time per iteration (ms): 41.7 | 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([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor:batch tensor:  tokens batch tensor:torch.Size([1, 1024])tokens
+  tokensbatch tensor:  labels torch.Size([1, 1024])
+torch.Size([1, 1024])batch tensor:torch.Size([1, 1024])
+
+ loss_mask batch tensor:batch tensor:torch.Size([1, 1024])  
+labelslabelsbatch tensor:   torch.Size([1, 1024])torch.Size([1, 1024])attention_mask
+
+ batch tensor:batch tensor:torch.Size([1, 1, 1024, 1024]) 
+ loss_maskloss_maskbatch tensor:   torch.Size([1, 1024])torch.Size([1, 1024])position_ids
+
+ batch tensor:batch tensor:torch.Size([1, 1024])  
+attention_maskattention_mask  torch.Size([1, 1, 1024, 1024])torch.Size([1, 1, 1024, 1024])
+
+batch tensor:batch tensor:  batch tensor:position_idsposition_ids   torch.Size([1, 1024])torch.Size([1, 1024])tokens
+
+ torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp:batch tensor after cp:  loss_masktokens  torch.Size([1, 128])
+torch.Size([1, 128])batch tensor after cp:
+ attention_maskbatch tensor after cp:  labelstorch.Size([1, 1, 128, 1024]) 
+torch.Size([1, 128])batch tensor after cp:
+ batch tensor after cp:position_ids  loss_masktorch.Size([1, 128]) 
+torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: batch tensor after cp:position_ids  torch.Size([1, 128])tokens
+ batch tensor after cp: torch.Size([1, 128])tokens
+ batch tensor after cp:torch.Size([1, 128]) 
+labels batch tensor after cp:torch.Size([1, 128]) 
+labelsbatch tensor after cp:  torch.Size([1, 128])loss_mask
+ batch tensor after cp:torch.Size([1, 128]) 
+loss_mask batch tensor after cp:torch.Size([1, 128]) 
+attention_mask batch tensor after cp:torch.Size([1, 1, 128, 1024]) 
+attention_maskbatch tensor after cp:  position_idstorch.Size([1, 1, 128, 1024]) 
+torch.Size([1, 128])batch tensor after cp:
+ position_ids torch.Size([1, 128])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor: tokens torch.Size([1, 1024])
+batch tensor: labels torch.Size([1, 1024])
+batch tensor: loss_mask torch.Size([1, 1024])
+batch tensor: attention_mask torch.Size([1, 1, 1024, 1024])
+batch tensor: position_ids torch.Size([1, 1024])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+batch tensor after cp: tokens torch.Size([1, 128])
+batch tensor after cp: labels torch.Size([1, 128])
+batch tensor after cp: loss_mask torch.Size([1, 128])
+batch tensor after cp: attention_mask torch.Size([1, 1, 128, 1024])
+batch tensor after cp: position_ids torch.Size([1, 128])
+Start exporting trace 9
+Done exporting trace 9
+ [2025-06-21 22:06:58] iteration       10/      10 | consumed samples:           10 | elapsed time per iteration (ms): 45.5 | 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 22:06:58 
+saving checkpoint at iteration      10 to gpt-checkpoint in torch_dist format
+DEBUG:megatron.training.checkpointing:rank: 5, takes 0.03761911392211914 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 0, takes 0.03771018981933594 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 2, takes 0.03774237632751465 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 1, takes 0.037798404693603516 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 3, takes 0.03775763511657715 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 6, takes 0.037831783294677734 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 4, takes 0.03794074058532715 to prepare state dict for ckpt 
+DEBUG:megatron.training.checkpointing:rank: 7, takes 0.037969350814819336 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