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

Bump transformers version #724

Merged
merged 16 commits into from
May 29, 2024
Merged

Bump transformers version #724

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
b84ef4a
Bump transformers version
echarlaix May 22, 2024
f68b2c7
fix default ignored scope for model using sdpa by default
echarlaix May 22, 2024
d120a45
fix quant ops test
echarlaix May 22, 2024
822606b
fix test
echarlaix May 22, 2024
b9c3718
Merge branch 'main' into bump
echarlaix May 23, 2024
1456ea2
fix test
echarlaix May 23, 2024
87903bc
fix ops
echarlaix May 23, 2024
451fc0f
tmp for test
echarlaix May 23, 2024
67598c1
style
echarlaix May 23, 2024
f3be817
update setup
echarlaix May 23, 2024
ae1f449
add ops quant num gpt2
echarlaix May 23, 2024
5fd2865
merge main in branch
echarlaix May 28, 2024
8ea75f7
tmp install optimum from source
echarlaix May 28, 2024
2163bf5
Merge branch 'main' into bump
echarlaix May 28, 2024
bd777cb
fix expected ops quant in test
echarlaix May 29, 2024
061bf47
update optimum version
echarlaix May 29, 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: 0 additions & 2 deletions examples/openvino/image-classification/configs/swin-base-jpqd.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,8 +36,6 @@
"ignored_scopes": [
"{re}.*__add___[0-1]",
"{re}.*layer_norm_0",
"{re}.*matmul_1",
"{re}.*__truediv__*"
]
}
]
2 changes: 0 additions & 2 deletions examples/openvino/question-answering/configs/bert-base-jpqd.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,8 +36,6 @@
"ignored_scopes": [
"{re}.*__add___[0-1]",
"{re}.*layer_norm_0",
"{re}.*matmul_1",
"{re}.*__truediv__*"
]
}
]
2 changes: 0 additions & 2 deletions examples/openvino/text-classification/configs/bert-base-jpqd.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,8 +40,6 @@
"ignored_scopes": [
"{re}.*__add___[0-1]",
"{re}.*layer_norm_0",
"{re}.*matmul_1",
"{re}.*__truediv__*"
]
}
]
2 changes: 0 additions & 2 deletions optimum/intel/openvino/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,8 +153,6 @@
"{re}.*Embedding.*",
"{re}.*add___.*",
"{re}.*layer_norm_.*",
"{re}.*matmul_1",
"{re}.*__truediv__.*",
],
}

Expand Down
4 changes: 2 additions & 2 deletions setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,8 +28,8 @@

INSTALL_REQUIRE = [
"torch>=1.11",
"transformers>=4.36.0,<4.41.0",
"optimum~=1.19",
"transformers>=4.36.0,<4.42.0",
"optimum~=1.20",
"datasets>=1.4.0",
"sentencepiece",
"scipy",
Expand Down
2 changes: 1 addition & 1 deletion tests/openvino/test_modeling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1679,7 +1679,7 @@ def test_compare_output_attentions(self, model_arch):
preprocessor = AutoFeatureExtractor.from_pretrained(model_id)
inputs = preprocessor(images=image, return_tensors="pt")

transformers_model = AutoModelForImageClassification.from_pretrained(model_id)
transformers_model = AutoModelForImageClassification.from_pretrained(model_id, attn_implementation="eager")
transformers_model.eval()
with torch.no_grad():
transformers_outputs = transformers_model(**inputs, output_attentions=True)
Expand Down
4 changes: 2 additions & 2 deletions tests/openvino/test_quantization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@

class OVQuantizerTest(unittest.TestCase):
SUPPORTED_ARCHITECTURES_TORCH_MODEL = (
(OVModelForSequenceClassification, "bert", 32, 35),
(OVModelForSequenceClassification, "bert", 22, 35),
(OVModelForCausalLM, "gpt2", 41, 3),
)
SUPPORTED_ARCHITECTURES_OV_MODEL = (
Expand Down Expand Up @@ -665,7 +665,7 @@ def preprocess_function(examples, tokenizer):


class OVTrainerTest(unittest.TestCase):
SUPPORTED_ARCHITECTURES_WITH_EXPECTED_QUANTIZED_MATMULS = (("distilbert-base-uncased", 49, 38),)
SUPPORTED_ARCHITECTURES_WITH_EXPECTED_QUANTIZED_MATMULS = (("distilbert-base-uncased", 67, 38),)

@parameterized.expand(SUPPORTED_ARCHITECTURES_WITH_EXPECTED_QUANTIZED_MATMULS)
def test_aware_training_quantization(self, model_name, expected_fake_quantize, expected_int8):
Expand Down
38 changes: 20 additions & 18 deletions tests/openvino/test_training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -322,30 +322,30 @@ def tearDown(self):
"default_quantization": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=DEFAULT_QUANTIZATION_CONFIG,
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
compression_metrics=["compression_loss"],
),
"distillation,default_quantization": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
teacher_model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=DEFAULT_QUANTIZATION_CONFIG,
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
),
"customized_quantization": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=CUSTOMIZED_QUANTIZATION_CONFIG,
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
compression_metrics=["compression_loss"],
),
"distillation,customized_quantization": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
teacher_model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=CUSTOMIZED_QUANTIZATION_CONFIG,
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
),
Expand All @@ -365,7 +365,7 @@ def tearDown(self):
"default_quantization,structured_movement_sparsity": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=[DEFAULT_QUANTIZATION_CONFIG, STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss"],
Expand All @@ -376,7 +376,7 @@ def tearDown(self):
CUSTOMIZED_QUANTIZATION_CONFIG,
STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT,
],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss"],
Expand All @@ -385,7 +385,7 @@ def tearDown(self):
model_id="hf-internal-testing/tiny-random-bert",
teacher_model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=[DEFAULT_QUANTIZATION_CONFIG, STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand All @@ -397,7 +397,7 @@ def tearDown(self):
CUSTOMIZED_QUANTIZATION_CONFIG,
STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT,
],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand All @@ -418,7 +418,7 @@ def tearDown(self):
"default_quantization,unstructured_movement_sparsity": OVTrainerTestDescriptor(
model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=[DEFAULT_QUANTIZATION_CONFIG, UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss"],
Expand All @@ -429,7 +429,7 @@ def tearDown(self):
CUSTOMIZED_QUANTIZATION_CONFIG,
UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT,
],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss"],
Expand All @@ -438,7 +438,7 @@ def tearDown(self):
model_id="hf-internal-testing/tiny-random-bert",
teacher_model_id="hf-internal-testing/tiny-random-bert",
nncf_compression_config=[DEFAULT_QUANTIZATION_CONFIG, UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand All @@ -450,7 +450,7 @@ def tearDown(self):
CUSTOMIZED_QUANTIZATION_CONFIG,
UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_BERT,
],
expected_fake_quantize=34,
expected_fake_quantize=22,
expected_int8=32,
expected_binary_masks=60,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand Down Expand Up @@ -553,7 +553,7 @@ def check_ovmodel_reshaping(self, ovmodel: OVModel):
"default_quantization": OVTrainerTestDescriptor(
model_id="yujiepan/tiny-random-swin-patch4-window7-224",
nncf_compression_config=DEFAULT_QUANTIZATION_CONFIG,
expected_fake_quantize=28,
expected_fake_quantize=36,
expected_int8=28,
compression_metrics=["compression_loss"],
),
Expand All @@ -572,15 +572,15 @@ def check_ovmodel_reshaping(self, ovmodel: OVModel):
"default_quantization,structured_movement_sparsity": OVTrainerTestDescriptor(
model_id="yujiepan/tiny-random-swin-patch4-window7-224",
nncf_compression_config=[STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_SWIN, DEFAULT_QUANTIZATION_CONFIG],
expected_fake_quantize=28,
expected_fake_quantize=36,
expected_int8=28,
expected_binary_masks=48,
compression_metrics=["compression_loss"],
),
"default_quantization,unstructured_movement_sparsity": OVTrainerTestDescriptor(
model_id="yujiepan/tiny-random-swin-patch4-window7-224",
nncf_compression_config=[UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_SWIN, DEFAULT_QUANTIZATION_CONFIG],
expected_fake_quantize=28,
expected_fake_quantize=36,
expected_int8=28,
expected_binary_masks=48,
compression_metrics=["compression_loss"],
Expand All @@ -589,7 +589,7 @@ def check_ovmodel_reshaping(self, ovmodel: OVModel):
model_id="yujiepan/tiny-random-swin-patch4-window7-224",
teacher_model_id="yujiepan/tiny-random-swin-patch4-window7-224",
nncf_compression_config=[STRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_SWIN, DEFAULT_QUANTIZATION_CONFIG],
expected_fake_quantize=28,
expected_fake_quantize=36,
expected_int8=28,
expected_binary_masks=48,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand All @@ -598,7 +598,7 @@ def check_ovmodel_reshaping(self, ovmodel: OVModel):
model_id="yujiepan/tiny-random-swin-patch4-window7-224",
teacher_model_id="yujiepan/tiny-random-swin-patch4-window7-224",
nncf_compression_config=[UNSTRUCTURED_MOVEMENT_SPARSITY_CONFIG_FOR_SWIN, DEFAULT_QUANTIZATION_CONFIG],
expected_fake_quantize=28,
expected_fake_quantize=36,
expected_int8=28,
expected_binary_masks=48,
compression_metrics=["compression_loss", "distillation_loss", "task_loss"],
Expand Down Expand Up @@ -797,7 +797,9 @@ def prepare_model_and_dataset(self, desc: OVTrainerTestDescriptor):

self.feature_extractor = AutoFeatureExtractor.from_pretrained(desc.model_id)
self.tokenizer = self.feature_extractor
self.model = AutoModelForAudioClassification.from_pretrained(desc.model_id, num_labels=self.num_labels)
self.model = AutoModelForAudioClassification.from_pretrained(
desc.model_id, num_labels=self.num_labels, attn_implementation="eager"
)
self.teacher_model = None
if desc.teacher_model_id:
self.teacher_model = AutoModelForAudioClassification.from_pretrained(
Expand Down
Loading