transformers
da971b22 - Keep relevant weights in fp32 when `model._keep_in_fp32_modules` is set even when `accelerate` is not installed (#26225)

Commit

2 years ago

Keep relevant weights in fp32 when `model._keep_in_fp32_modules` is set even when `accelerate` is not installed (#26225) * fix bug where weight would not be kept in fp32 * nit * address review comments * fix test

References

#26225 - Keep relevant weights in fp32 when `model._keep_in_fp32_modules` is set even when `accelerate` is not installed

Author

fxmarty

Parents

e3a4bd2b

transformers da971b22 - Keep relevant weights in fp32 when `model._keep_in_fp32_modules` is set even when `accelerate` is not installed (#26225)

transformers
da971b22 - Keep relevant weights in fp32 when `model._keep_in_fp32_modules` is set even when `accelerate` is not installed (#26225)