Skip to content

Disable inductor config setter by default #36608

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
ArthurZucker merged 8 commits into from
Mar 20, 2025
Merged

Disable inductor config setter by default #36608

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
688f56e
Disable inductor config setter by default
HDCharles Mar 7, 2025
51b5d96
remove default settings in autoquant too
HDCharles Mar 7, 2025
2191025
Add info to torchao.md about recommended settings
HDCharles Mar 7, 2025
ebcb6b7
satisfying Ruff format
HDCharles Mar 7, 2025
d688f78
Merge branch 'main' into default_no_inductor_setter
HDCharles Mar 10, 2025
677d96f
Merge branch 'main' into default_no_inductor_setter
SunMarc Mar 13, 2025
41820d3
Merge branch 'main' into default_no_inductor_setter
SunMarc Mar 13, 2025
b63e7fb
Merge branch 'main' into default_no_inductor_setter
HDCharles Mar 19, 2025
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
3 changes: 3 additions & 0 deletions docs/source/en/quantization/torchao.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,9 @@ output = bf16_model.generate(**input_ids, max_new_tokens=10, cache_implementatio
print("bf16 model:", benchmark_fn(bf16_model.generate, **input_ids, max_new_tokens=MAX_NEW_TOKENS, cache_implementation="static"))
```

> [!TIP]
> For best performance, you can use recommended settings by calling `torchao.quantization.utils.recommended_inductor_config_setter()`

</hfoption>
<hfoption id="automatic">

Expand Down
4 changes: 2 additions & 2 deletions src/transformers/quantizers/quantizer_torchao.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,7 +203,7 @@ def create_quantized_param(
module.extra_repr = types.MethodType(_linear_extra_repr, module)
else:
module._parameters[tensor_name] = torch.nn.Parameter(param_value).to(device=target_device)
quantize_(module, self.quantization_config.get_apply_tensor_subclass())
quantize_(module, self.quantization_config.get_apply_tensor_subclass(), set_inductor_config=False)

def _process_model_after_weight_loading(self, model, **kwargs):
"""No process required for torchao quantized model"""
Expand All @@ -213,7 +213,7 @@ def _process_model_after_weight_loading(self, model, **kwargs):

model = torch.compile(model, mode="max-autotune")
model = autoquant(
model, qtensor_class_list=ALL_AUTOQUANT_CLASS_LIST, **self.quantization_config.quant_type_kwargs
model, qtensor_class_list=ALL_AUTOQUANT_CLASS_LIST, set_inductor_config=False, **self.quantization_config.quant_type_kwargs
)
return model
return
Expand Down