Skip to content

Conversation

@KeremTurgutlu
Copy link

B.dtype which is the data type of the packed 4bit values can now have different dtypes other than uint8, so the check should consider it.

bnb_quantized is missing when Params4bit.__new__() used for setting a param, so added it as an arg.

module._parameters[name] = param_cls(module._parameters[name].to(device), **kwargs)

These fix multiple tests in peft/tests/test_common_gpu.py

@warner-benjamin
Copy link

Looks good. Thanks.

@warner-benjamin warner-benjamin merged commit 754e75b into AnswerDotAI:cuda_fix_quant_storage_dtype Jan 9, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants