Skip to content

AssertionError: model parallel group is not initialized #104

@nikich340

Description

@nikich340
Traceback (most recent call last):
  File "D:\_ruGPT3\ru-gpts\run_gpt3medium.py", line 11, in <module>
    gpt = RuGPT3XL.from_pretrained("sber-gpt3-medium", seq_len=512)
  File "D:\_ruGPT3\ru-gpts\src\xl_wrapper.py", line 178, in from_pretrained
    mpu.model_parallel_cuda_manual_seed(seed)
  File "D:\_ruGPT3\ru-gpts\src\mpu\random.py", line 217, in model_parallel_cuda_manual_seed
    model_parallel_seed = offset + get_model_parallel_rank()
  File "D:\_ruGPT3\ru-gpts\src\mpu\initialize.py", line 109, in get_model_parallel_rank
    return torch.distributed.get_rank(group=get_model_parallel_group())
  File "D:\_ruGPT3\ru-gpts\src\mpu\initialize.py", line 90, in get_model_parallel_group
    assert _MODEL_PARALLEL_GROUP is not None, \
AssertionError: model parallel group is not initialized

Used "Usage" example code but with rugpt3-medium model. Os: windows 10, py 3.9, gpu: rtx 3060 laptop.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions