Skip to content
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

Jump to bottom

Enhance quantization code for neural_compressor #309

Merged
merged 1 commit into from
May 12, 2023
Merged

Enhance quantization code for neural_compressor #309

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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: 1 addition & 1 deletion optimum/intel/neural_compressor/configuration.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@

from typing import Dict, Optional, Union

from neural_compressor.conf.pythonic_config import DistillationConfig, WeightPruningConfig, _BaseQuantizationConfig
from neural_compressor.config import DistillationConfig, WeightPruningConfig, _BaseQuantizationConfig

from optimum.configuration_utils import BaseConfig

Expand Down
26 changes: 17 additions & 9 deletions optimum/intel/neural_compressor/quantization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -165,16 +165,21 @@ def quantize(
calibration_dataloader = None

if INCQuantizationMode(quantization_config.approach) == INCQuantizationMode.STATIC:
# Since PyTorch fx trace does not really require an example_inputs, only need calibration_dataset or calibration_fn here.
if calibration_dataset is None and self.calibration_fn is None:
raise ValueError(
"Post-training static quantization needs a calibration dataset or a calibration_function."
)
if calibration_dataset is None:
raise ValueError("Post-training static quantization needs a calibration dataset.")

quantization_config.calibration_sampling_size = len(calibration_dataset)
calibration_dataloader = self._get_calibration_dataloader(
calibration_dataset=calibration_dataset,
batch_size=batch_size,
remove_unused_columns=remove_unused_columns,
data_collator=data_collator,
)
calibration_dataloader = None
else:
quantization_config.calibration_sampling_size = len(calibration_dataset)
calibration_dataloader = self._get_calibration_dataloader(
calibration_dataset=calibration_dataset,
batch_size=batch_size,
remove_unused_columns=remove_unused_columns,
data_collator=data_collator,
)

if isinstance(self._original_model.config, PretrainedConfig):
self._original_model.config.backend = quantization_config.backend
Expand All @@ -193,7 +198,10 @@ def quantize(
" accuracy tolerance has been found. Either the tolerance or the number of trials need to be increased."
)
if isinstance(self._original_model.config, PretrainedConfig):
original_dtype = self._original_model.config.torch_dtype
self._original_model.config.torch_dtype = "int8"
self._original_model.config.save_pretrained(save_directory)
self._original_model.config.torch_dtype = original_dtype

self._quantized_model = compressed_model._model

Expand Down
Loading