Skip to content

Use torch.device instead of current device index for BnB quantizer #10069

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Dec 5, 2024
Merged

Use torch.device instead of current device index for BnB quantizer #10069

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
aeeadc7
update
a-r-r-o-w Dec 1, 2024
7261559
Merge branch 'main' into param-device-bnb
a-r-r-o-w Dec 4, 2024
126d84e
apply review suggestion
a-r-r-o-w Dec 4, 2024
1a171c3
Merge branch 'main' into param-device-bnb
sayakpaul Dec 5, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/diffusers/models/model_loading_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -176,6 +176,8 @@ def load_model_dict_into_meta(
hf_quantizer=None,
keep_in_fp32_modules=None,
) -> List[str]:
if device is not None and not isinstance(device, (str, torch.device)):
raise ValueError(f"Expected device to have type `str` or `torch.device`, but got {type(device)=}.")
if hf_quantizer is None:
device = device or torch.device("cpu")
dtype = dtype or torch.float32
Expand Down
2 changes: 1 addition & 1 deletion src/diffusers/models/modeling_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -836,7 +836,7 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
param_device = "cpu"
# TODO (sayakpaul, SunMarc): remove this after model loading refactor
elif is_quant_method_bnb:
param_device = torch.cuda.current_device()
param_device = torch.device(torch.cuda.current_device())
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good to me!

@yiyixuxu WDYT?

Cc: @SunMarc as well.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

let's throw an error in load_model_dict_into_meta when device is passed as index??

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Throws a value error now. @yiyixuxu

@sayakpaul, the integration tests pass:

(nightly-venv) (nightly-venv) aryan@hf-dgx-01:~/work/diffusers$ RUN_SLOW=1 CUDA_VISIBLE_DEVICES="3" pytest -s tests/quantization/bnb/test_4bit.py::SlowBnb4BitFluxTests
========================================================================================================================================= test session starts ==========================================================================================================================================
platform linux -- Python 3.10.14, pytest-8.3.2, pluggy-1.5.0
rootdir: /home/aryan/work/diffusers
configfile: pyproject.toml
plugins: timeout-2.3.1, requests-mock-1.10.0, xdist-3.6.1, anyio-4.6.2.post1
collected 1 item                                                                                                                                                                                                                                                                                       

tests/quantization/bnb/test_4bit.py Unused kwargs: ['_load_in_4bit', '_load_in_8bit', 'quant_method']. These kwargs are not used in <class 'transformers.utils.quantization_config.BitsAndBytesConfig'>.
`low_cpu_mem_usage` was None, now default to True since model is quantized.
Loading pipeline components...:  14%|████████████████████████████████▋                                                                                                                                                                                                    | 1/7 [00:00<00:00,  9.00it/s]You set `add_prefix_space`. The tokenizer needs to be converted from the slow tokenizers
Loading pipeline components...: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00,  9.26it/s]
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:14<00:00,  1.50s/it]
.

========================================================================================================================================== 1 passed in 52.27s ===================================================================

state_dict = load_state_dict(model_file, variant=variant)
model._convert_deprecated_attention_blocks(state_dict)

Expand Down
Loading