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

Deprecate TF + JAX (#38758)

Browse Source 
* Scatter deprecation warnings around

* Delete the tests

* Make logging work properly!
This commit is contained in:
Matt
2025-06-11 17:28:06 +01:00
committed by GitHub
parent 337757cbd5
commit 9f563ada70
121 changed files with 27 additions and 50030 deletions
Download Patch File Download Diff File Expand all files Collapse all files

102
tests/models/auto/test_modeling_flax_auto.py
View File

@@ -1,102 +0,0 @@
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
from transformers import AutoConfig, AutoTokenizer, BertConfig, TensorType, is_flax_available
from transformers.testing_utils import DUMMY_UNKNOWN_IDENTIFIER, require_flax, slow
if is_flax_available():
import jax
from transformers.models.auto.modeling_flax_auto import FlaxAutoModel
from transformers.models.bert.modeling_flax_bert import FlaxBertModel
from transformers.models.roberta.modeling_flax_roberta import FlaxRobertaModel
@require_flax
class FlaxAutoModelTest(unittest.TestCase):
@slow
def test_bert_from_pretrained(self):
for model_name in ["google-bert/bert-base-cased", "google-bert/bert-large-uncased"]:
with self.subTest(model_name):
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = FlaxAutoModel.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, FlaxBertModel)
@slow
def test_roberta_from_pretrained(self):
for model_name in ["FacebookAI/roberta-base", "FacebookAI/roberta-large"]:
with self.subTest(model_name):
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = FlaxAutoModel.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, FlaxRobertaModel)
@slow
def test_bert_jax_jit(self):
for model_name in ["google-bert/bert-base-cased", "google-bert/bert-large-uncased"]:
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = FlaxBertModel.from_pretrained(model_name)
tokens = tokenizer("Do you support jax jitted function?", return_tensors=TensorType.JAX)
@jax.jit
def eval(**kwargs):
return model(**kwargs)
eval(**tokens).block_until_ready()
@slow
def test_roberta_jax_jit(self):
for model_name in ["FacebookAI/roberta-base", "FacebookAI/roberta-large"]:
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = FlaxRobertaModel.from_pretrained(model_name)
tokens = tokenizer("Do you support jax jitted function?", return_tensors=TensorType.JAX)
@jax.jit
def eval(**kwargs):
return model(**kwargs)
eval(**tokens).block_until_ready()
def test_repo_not_found(self):
with self.assertRaisesRegex(
EnvironmentError, "bert-base is not a local folder and is not a valid model identifier"
):
_ = FlaxAutoModel.from_pretrained("bert-base")
def test_revision_not_found(self):
with self.assertRaisesRegex(
EnvironmentError, r"aaaaaa is not a valid git identifier \(branch name, tag name or commit id\)"
):
_ = FlaxAutoModel.from_pretrained(DUMMY_UNKNOWN_IDENTIFIER, revision="aaaaaa")
def test_model_file_not_found(self):
with self.assertRaisesRegex(
EnvironmentError,
"hf-internal-testing/config-no-model does not appear to have a file named flax_model.msgpack",
):
_ = FlaxAutoModel.from_pretrained("hf-internal-testing/config-no-model")
def test_model_from_pt_suggestion(self):
with self.assertRaisesRegex(EnvironmentError, "Use `from_pt=True` to load this model"):
_ = FlaxAutoModel.from_pretrained("hf-internal-testing/tiny-bert-pt-only")

310
tests/models/auto/test_modeling_tf_auto.py
View File

@@ -1,310 +0,0 @@
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
import copy
import tempfile
import unittest
from transformers import CONFIG_MAPPING, AutoConfig, BertConfig, GPT2Config, T5Config, TapasConfig, is_tf_available
from transformers.testing_utils import (
DUMMY_UNKNOWN_IDENTIFIER,
SMALL_MODEL_IDENTIFIER,
RequestCounter,
require_tensorflow_probability,
require_tf,
slow,
)
from ..bert.test_modeling_bert import BertModelTester
if is_tf_available():
from transformers import (
TFAutoModel,
TFAutoModelForCausalLM,
TFAutoModelForMaskedLM,
TFAutoModelForPreTraining,
TFAutoModelForQuestionAnswering,
TFAutoModelForSeq2SeqLM,
TFAutoModelForSequenceClassification,
TFAutoModelForTableQuestionAnswering,
TFAutoModelForTokenClassification,
TFAutoModelWithLMHead,
TFBertForMaskedLM,
TFBertForPreTraining,
TFBertForQuestionAnswering,
TFBertForSequenceClassification,
TFBertModel,
TFFunnelBaseModel,
TFFunnelModel,
TFGPT2LMHeadModel,
TFRobertaForMaskedLM,
TFT5ForConditionalGeneration,
TFTapasForQuestionAnswering,
)
from transformers.models.auto.modeling_tf_auto import (
TF_MODEL_FOR_CAUSAL_LM_MAPPING,
TF_MODEL_FOR_MASKED_LM_MAPPING,
TF_MODEL_FOR_PRETRAINING_MAPPING,
TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
TF_MODEL_MAPPING,
)
class NewModelConfig(BertConfig):
model_type = "new-model"
if is_tf_available():
class TFNewModel(TFBertModel):
config_class = NewModelConfig
@require_tf
class TFAutoModelTest(unittest.TestCase):
@slow
def test_model_from_pretrained(self):
model_name = "google-bert/bert-base-cased"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = TFAutoModel.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFBertModel)
@slow
def test_model_for_pretraining_from_pretrained(self):
model_name = "google-bert/bert-base-cased"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = TFAutoModelForPreTraining.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFBertForPreTraining)
@slow
def test_model_for_causal_lm(self):
model_name = "openai-community/gpt2"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, GPT2Config)
model = TFAutoModelForCausalLM.from_pretrained(model_name)
model, loading_info = TFAutoModelForCausalLM.from_pretrained(model_name, output_loading_info=True)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFGPT2LMHeadModel)
@slow
def test_lmhead_model_from_pretrained(self):
model_name = "openai-community/gpt2"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, GPT2Config)
model = TFAutoModelWithLMHead.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFGPT2LMHeadModel)
@slow
def test_model_for_masked_lm(self):
model_name = "google-bert/bert-base-uncased"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = TFAutoModelForMaskedLM.from_pretrained(model_name)
model, loading_info = TFAutoModelForMaskedLM.from_pretrained(model_name, output_loading_info=True)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFBertForMaskedLM)
@slow
def test_model_for_encoder_decoder_lm(self):
model_name = "google-t5/t5-base"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, T5Config)
model = TFAutoModelForSeq2SeqLM.from_pretrained(model_name)
model, loading_info = TFAutoModelForSeq2SeqLM.from_pretrained(model_name, output_loading_info=True)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFT5ForConditionalGeneration)
@slow
def test_sequence_classification_model_from_pretrained(self):
# model_name = 'openai-community/gpt2'
for model_name in ["google-bert/bert-base-uncased"]:
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = TFAutoModelForSequenceClassification.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFBertForSequenceClassification)
@slow
def test_question_answering_model_from_pretrained(self):
# model_name = 'openai-community/gpt2'
for model_name in ["google-bert/bert-base-uncased"]:
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, BertConfig)
model = TFAutoModelForQuestionAnswering.from_pretrained(model_name)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFBertForQuestionAnswering)
@slow
@require_tensorflow_probability
def test_table_question_answering_model_from_pretrained(self):
model_name = "google/tapas-base"
config = AutoConfig.from_pretrained(model_name)
self.assertIsNotNone(config)
self.assertIsInstance(config, TapasConfig)
model = TFAutoModelForTableQuestionAnswering.from_pretrained(model_name)
model, loading_info = TFAutoModelForTableQuestionAnswering.from_pretrained(
model_name, output_loading_info=True
)
self.assertIsNotNone(model)
self.assertIsInstance(model, TFTapasForQuestionAnswering)
def test_from_pretrained_identifier(self):
model = TFAutoModelWithLMHead.from_pretrained(SMALL_MODEL_IDENTIFIER)
self.assertIsInstance(model, TFBertForMaskedLM)
self.assertEqual(model.num_parameters(), 14410)
self.assertEqual(model.num_parameters(only_trainable=True), 14410)
def test_from_identifier_from_model_type(self):
model = TFAutoModelWithLMHead.from_pretrained(DUMMY_UNKNOWN_IDENTIFIER)
self.assertIsInstance(model, TFRobertaForMaskedLM)
self.assertEqual(model.num_parameters(), 14410)
self.assertEqual(model.num_parameters(only_trainable=True), 14410)
def test_from_pretrained_with_tuple_values(self):
# For the auto model mapping, FunnelConfig has two models: FunnelModel and FunnelBaseModel
model = TFAutoModel.from_pretrained("sgugger/funnel-random-tiny")
self.assertIsInstance(model, TFFunnelModel)
config = copy.deepcopy(model.config)
config.architectures = ["FunnelBaseModel"]
model = TFAutoModel.from_config(config)
model.build_in_name_scope()
self.assertIsInstance(model, TFFunnelBaseModel)
with tempfile.TemporaryDirectory() as tmp_dir:
model.save_pretrained(tmp_dir)
model = TFAutoModel.from_pretrained(tmp_dir)
self.assertIsInstance(model, TFFunnelBaseModel)
def test_new_model_registration(self):
try:
AutoConfig.register("new-model", NewModelConfig)
auto_classes = [
TFAutoModel,
TFAutoModelForCausalLM,
TFAutoModelForMaskedLM,
TFAutoModelForPreTraining,
TFAutoModelForQuestionAnswering,
TFAutoModelForSequenceClassification,
TFAutoModelForTokenClassification,
]
for auto_class in auto_classes:
with self.subTest(auto_class.__name__):
# Wrong config class will raise an error
with self.assertRaises(ValueError):
auto_class.register(BertConfig, TFNewModel)
auto_class.register(NewModelConfig, TFNewModel)
# Trying to register something existing in the Transformers library will raise an error
with self.assertRaises(ValueError):
auto_class.register(BertConfig, TFBertModel)
# Now that the config is registered, it can be used as any other config with the auto-API
tiny_config = BertModelTester(self).get_config()
config = NewModelConfig(**tiny_config.to_dict())
model = auto_class.from_config(config)
model.build_in_name_scope()
self.assertIsInstance(model, TFNewModel)
with tempfile.TemporaryDirectory() as tmp_dir:
model.save_pretrained(tmp_dir)
new_model = auto_class.from_pretrained(tmp_dir)
self.assertIsInstance(new_model, TFNewModel)
finally:
if "new-model" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["new-model"]
for mapping in (
TF_MODEL_MAPPING,
TF_MODEL_FOR_PRETRAINING_MAPPING,
TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
TF_MODEL_FOR_CAUSAL_LM_MAPPING,
TF_MODEL_FOR_MASKED_LM_MAPPING,
):
if NewModelConfig in mapping._extra_content:
del mapping._extra_content[NewModelConfig]
def test_repo_not_found(self):
with self.assertRaisesRegex(
EnvironmentError, "bert-base is not a local folder and is not a valid model identifier"
):
_ = TFAutoModel.from_pretrained("bert-base")
def test_revision_not_found(self):
with self.assertRaisesRegex(
EnvironmentError, r"aaaaaa is not a valid git identifier \(branch name, tag name or commit id\)"
):
_ = TFAutoModel.from_pretrained(DUMMY_UNKNOWN_IDENTIFIER, revision="aaaaaa")
def test_model_file_not_found(self):
with self.assertRaisesRegex(
EnvironmentError,
"hf-internal-testing/config-no-model does not appear to have a file named pytorch_model.bin",
):
_ = TFAutoModel.from_pretrained("hf-internal-testing/config-no-model")
def test_model_from_pt_suggestion(self):
with self.assertRaisesRegex(EnvironmentError, "Use `from_pt=True` to load this model"):
_ = TFAutoModel.from_pretrained("hf-internal-testing/tiny-bert-pt-only")
@unittest.skip("Failing on main")
def test_cached_model_has_minimum_calls_to_head(self):
# Make sure we have cached the model.
_ = TFAutoModel.from_pretrained("hf-internal-testing/tiny-random-bert")
with RequestCounter() as counter:
_ = TFAutoModel.from_pretrained("hf-internal-testing/tiny-random-bert")
self.assertEqual(counter["GET"], 0)
self.assertEqual(counter["HEAD"], 1)
self.assertEqual(counter.total_calls, 1)
# With a sharded checkpoint
_ = TFAutoModel.from_pretrained("ArthurZ/tiny-random-bert-sharded")
with RequestCounter() as counter:
_ = TFAutoModel.from_pretrained("ArthurZ/tiny-random-bert-sharded")
self.assertEqual(counter["GET"], 0)
self.assertEqual(counter["HEAD"], 1)
self.assertEqual(counter.total_calls, 1)