Watchdog caught collective operation timeout

[minji-o-j]

27 Jul 06:20    INFO Soft link created: saved/PTG-dd-2023-Jul-27_01-10-24/checkpoint_best -> /workspace/TextBox/saved/PTG-dd-2023-Jul-27_01-10-24/checkpoint_epoch-5
27 Jul 06:20    INFO ====== Finished training, best validation result at train epoch 5 ======
27 Jul 06:20    INFO Best valid result: score: 65.06, <bleu-1: 33.65>, <bleu-2: 31.41>, bleu-3: 33.47, bleu-4: 32.60, distinct-1: 3.13, distinct-2: 12.90, distinct-3: 20.85, distinct-4: 26.43
27 Jul 06:20    INFO Loading model structure and parameters from saved/PTG-dd-2023-Jul-27_01-10-24/checkpoint_best ...

[E ProcessGroupNCCL.cpp:821] [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=19450, OpType=ALLGATHER, Timeout(ms)=1800000) ran for 1801798 milliseconds before timing out.
[E ProcessGroupNCCL.cpp:456] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
[E ProcessGroupNCCL.cpp:461] To avoid data inconsistency, we are taking the entire process down.
terminate called after throwing an instance of 'std::runtime_error'
  what():  [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=19450, OpType=ALLGATHER, Timeout(ms)=1800000) ran for 1801798 milliseconds before timing out.
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -6) local_rank: 0 (pid: 491) of binary: /opt/conda/bin/python
Traceback (most recent call last):
  File "/opt/conda/bin/accelerate", line 8, in <module>
    sys.exit(main())
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/accelerate_cli.py", line 45, in main
    args.func(args)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/launch.py", line 950, in launch_command
    multi_gpu_launcher(args)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/launch.py", line 642, in multi_gpu_launcher
    distrib_run.run(args)
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/run.py", line 753, in run
    elastic_launch(
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 246, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
====================================================
run_textbox.py FAILED
----------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
----------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2023-07-27_06:51:11
  host      : b94bc7c0de46
  rank      : 0 (local_rank: 0)
  exitcode  : -6 (pid: 491)
  error_file: <N/A>
  traceback : Signal 6 (SIGABRT) received by PID 491
====================================================

[minji-o-j]

WongKinYiu/yolov7#714

[minji-o-j]

export NCCL_P2P_LEVEL=NVL
cmd에 이걸 한번씩 입력하자 !

[minji-o-j]

01 Aug 11:14    INFO  Validation  6 [time: 537.22s, score: 84.19, <bleu-1: 45.70>, <bleu-2: 38.49>, bleu-3: 38.10, bleu-4: 35.58, distinct-1: 1.44, distinct-2: 7.43, distinct-3: 15.43, distinct-4: 24.47]
01 Aug 11:15    INFO Early stopped at 3 non-best validation.
01 Aug 11:15    INFO Soft link created: saved/PTG-pc-2023-Aug-01_03-24-10/checkpoint_best -> /workspace/TextBox/saved/PTG-pc-2023-Aug-01_03-24-10/checkpoint_epoch-3
01 Aug 11:15    INFO ====== Finished training, best validation result at train epoch 3 ======
01 Aug 11:15    INFO Best valid result: score: 87.05, <bleu-1: 48.09>, <bleu-2: 38.96>, bleu-3: 37.21, bleu-4: 33.89, distinct-1: 1.24, distinct-2: 6.65, distinct-3: 14.27, distinct-4: 22.85
01 Aug 11:15    INFO Loading model structure and parameters from saved/PTG-pc-2023-Aug-01_03-24-10/checkpoint_best ...
[E ProcessGroupNCCL.cpp:821] [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=183888, OpType=ALLGATHER, Timeout(ms)=5400000) ran for 5408715 milliseconds before timing out.
01 Aug 12:45    ERROR Traceback (most recent call last):
  File "/workspace/TextBox/textbox/utils/dashboard.py", line 316, in new_experiment
    yield True
  File "/workspace/TextBox/textbox/quick_start/experiment.py", line 130, in run
    self._do_test()
  File "/workspace/TextBox/textbox/quick_start/experiment.py", line 112, in _do_test
    self.test_result = self.trainer.evaluate(self.test_data, load_best_model=self.do_train)
  File "/opt/conda/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/workspace/TextBox/textbox/trainer/trainer.py", line 481, in evaluate
    self.model = self.accelerator.prepare(self.model)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/accelerator.py", line 1199, in prepare
    result = tuple(
  File "/opt/conda/lib/python3.9/site-packages/accelerate/accelerator.py", line 1200, in <genexpr>
    self._prepare_one(obj, first_pass=True, device_placement=d) for obj, d in zip(args, device_placement)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/accelerator.py", line 1027, in _prepare_one
    return self.prepare_model(obj, device_placement=device_placement)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/accelerator.py", line 1295, in prepare_model
    model = torch.nn.parallel.DistributedDataParallel(
  File "/opt/conda/lib/python3.9/site-packages/torch/nn/parallel/distributed.py", line 655, in __init__
    _verify_param_shape_across_processes(self.process_group, parameters)
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/utils.py", line 112, in _verify_param_shape_across_processes
    return dist._verify_params_across_processes(process_group, tensors, logger)
RuntimeError: DDP expects same model across all ranks, but Rank 0 has 521 params, while rank 1 has inconsistent 24 params.

[E ProcessGroupNCCL.cpp:456] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
[E ProcessGroupNCCL.cpp:461] To avoid data inconsistency, we are taking the entire process down.
terminate called after throwing an instance of 'std::runtime_error'
  what():  [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=183888, OpType=ALLGATHER, Timeout(ms)=5400000) ran for 5408715 milliseconds before timing out.
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -6) local_rank: 0 (pid: 423) of binary: /opt/conda/bin/python
Traceback (most recent call last):
  File "/opt/conda/bin/accelerate", line 8, in <module>
    sys.exit(main())
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/accelerate_cli.py", line 45, in main
    args.func(args)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/launch.py", line 950, in launch_command
    multi_gpu_launcher(args)
  File "/opt/conda/lib/python3.9/site-packages/accelerate/commands/launch.py", line 642, in multi_gpu_launcher
    distrib_run.run(args)
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/run.py", line 753, in run
    elastic_launch(
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/opt/conda/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 246, in launch_agent
    raise ChildFailedError(

[zmtttt]

hello！ have you solved the problem？ I met the same problem