Skip to content

[FIXED] compiled_autograd.enable() Gemma #1250

@InderjeetVishnoi

Description

@InderjeetVishnoi

I am simply running the gemma-2-2b example notebook of unsloth [https://colab.research.google.com/drive/1weTpKOjBZxZJ5PQ-Ql8i6ptAY2x-FWVA?usp=sharing] and getting the cited error.

Stacktrace of the error looks like below :

RuntimeError Traceback (most recent call last)
in <cell line: 1>()
----> 1 trainer_stats = trainer.train()

39 frames
/usr/local/lib/python3.10/dist-packages/torch/_dynamo/compiled_autograd.py in enable(compiler_fn)
497 @contextlib.contextmanager
498 def enable(compiler_fn):
--> 499 prior = torch._C._dynamo.compiled_autograd.set_autograd_compiler(
500 functools.partial(AutogradCompilerInstance, compiler_fn)
501 )

RuntimeError: compiled_autograd.enable() requires no threads in backwards()

I have started facing this error lately only, can you suggest what might be wrong.

pytorch version:2.5.0+cu121

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions