SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix CI by tweaking torchao tests (#34832)

Browse Source
This commit is contained in:
Marc Sun
2024-11-20 20:28:51 +01:00
committed by GitHub
parent bf42c3bd4b
commit 3cb8676a91
2 changed files with 13 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/transformers/utils/quantization_config.py
View File

@@ -1264,8 +1264,13 @@ class TorchAoConfig(QuantizationConfigMixin):
r""" r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values. Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
""" """
if is_torchao_available():
if not version.parse(importlib.metadata.version("torchao")) >= version.parse("0.4.0"): if not version.parse(importlib.metadata.version("torchao")) >= version.parse("0.4.0"):
raise ValueError("Requires torchao 0.4.0 version and above") raise ValueError("Requires torchao 0.4.0 version and above")
else:
raise ValueError(
"TorchAoConfig requires torchao to be installed, please install with `pip install torchao`"
)
_STR_TO_METHOD = self._get_torchao_quant_type_to_method() _STR_TO_METHOD = self._get_torchao_quant_type_to_method()
if self.quant_type not in _STR_TO_METHOD.keys(): if self.quant_type not in _STR_TO_METHOD.keys():

11
tests/quantization/torchao_integration/test_torchao.py
View File

@@ -246,12 +246,13 @@ class TorchAoSerializationTest(unittest.TestCase):
# TODO: investigate why we don't have the same output as the original model for this test # TODO: investigate why we don't have the same output as the original model for this test
SERIALIZED_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)" SERIALIZED_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)"
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0" model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
quant_config = TorchAoConfig("int4_weight_only", group_size=32) quant_scheme, quant_scheme_kwargs = "int4_weight_only", {"group_size": 32}
device = "cuda:0" device = "cuda:0"
# called only once for all test in this class # called only once for all test in this class
@classmethod @classmethod
def setUpClass(cls): def setUpClass(cls):
cls.quant_config = TorchAoConfig(cls.quant_scheme, **cls.quant_scheme_kwargs)
cls.quantized_model = AutoModelForCausalLM.from_pretrained( cls.quantized_model = AutoModelForCausalLM.from_pretrained(
cls.model_name, cls.model_name,
torch_dtype=torch.bfloat16, torch_dtype=torch.bfloat16,
@@ -290,21 +291,21 @@ class TorchAoSerializationTest(unittest.TestCase):
class TorchAoSerializationW8A8Test(TorchAoSerializationTest): class TorchAoSerializationW8A8Test(TorchAoSerializationTest):
quant_config = TorchAoConfig("int8_dynamic_activation_int8_weight") quant_scheme, quant_scheme_kwargs = "int8_dynamic_activation_int8_weight", {}
ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)" ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)"
SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT
device = "cuda:0" device = "cuda:0"
class TorchAoSerializationW8Test(TorchAoSerializationTest): class TorchAoSerializationW8Test(TorchAoSerializationTest):
quant_config = TorchAoConfig("int8_weight_only") quant_scheme, quant_scheme_kwargs = "int8_weight_only", {}
ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)" ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)"
SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT
device = "cuda:0" device = "cuda:0"
class TorchAoSerializationW8A8CPUTest(TorchAoSerializationTest): class TorchAoSerializationW8A8CPUTest(TorchAoSerializationTest):
quant_config = TorchAoConfig("int8_dynamic_activation_int8_weight") quant_scheme, quant_scheme_kwargs = "int8_dynamic_activation_int8_weight", {}
ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)" ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)"
SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT
device = "cpu" device = "cpu"
@@ -318,7 +319,7 @@ class TorchAoSerializationW8A8CPUTest(TorchAoSerializationTest):
class TorchAoSerializationW8CPUTest(TorchAoSerializationTest): class TorchAoSerializationW8CPUTest(TorchAoSerializationTest):
quant_config = TorchAoConfig("int8_weight_only") quant_scheme, quant_scheme_kwargs = "int8_weight_only", {}
ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)" ORIGINAL_EXPECTED_OUTPUT = "What are we having for dinner?\n\nJessica: (smiling)"
SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT SERIALIZED_EXPECTED_OUTPUT = ORIGINAL_EXPECTED_OUTPUT
device = "cpu" device = "cpu"