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

[tests] remove pt_tf equivalence tests (#36253)

Browse Source
This commit is contained in:
Joao Gante
2025-02-19 11:55:11 +00:00
committed by GitHub
parent 1a81d774b1
commit 0863eef248
60 changed files with 56 additions and 2438 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
tests/utils/test_add_new_model_like.py
View File

@@ -736,55 +736,6 @@ NEW_BERT_CONSTANT = "value"
self.assertIsNone(bert_model_patterns.feature_extractor_class)
self.assertIsNone(bert_model_patterns.processor_class)
def test_retrieve_info_for_model_pt_tf_with_bert(self):
bert_info = retrieve_info_for_model("bert", frameworks=["pt", "tf"])
bert_classes = [
"BertForTokenClassification",
"BertForQuestionAnswering",
"BertForNextSentencePrediction",
"BertForSequenceClassification",
"BertForMaskedLM",
"BertForMultipleChoice",
"BertModel",
"BertForPreTraining",
"BertLMHeadModel",
]
expected_model_classes = {"pt": set(bert_classes), "tf": {f"TF{m}" for m in bert_classes}}
self.assertEqual(set(bert_info["frameworks"]), {"pt", "tf"})
model_classes = {k: set(v) for k, v in bert_info["model_classes"].items()}
self.assertEqual(model_classes, expected_model_classes)
all_bert_files = bert_info["model_files"]
model_files = {str(Path(f).relative_to(REPO_PATH)) for f in all_bert_files["model_files"]}
bert_model_files = BERT_MODEL_FILES - {"src/transformers/models/bert/modeling_flax_bert.py"}
self.assertEqual(model_files, bert_model_files)
test_files = {str(Path(f).relative_to(REPO_PATH)) for f in all_bert_files["test_files"]}
bert_test_files = {
"tests/models/bert/test_tokenization_bert.py",
"tests/models/bert/test_modeling_bert.py",
"tests/models/bert/test_modeling_tf_bert.py",
}
self.assertEqual(test_files, bert_test_files)
doc_file = str(Path(all_bert_files["doc_file"]).relative_to(REPO_PATH))
self.assertEqual(doc_file, "docs/source/en/model_doc/bert.md")
self.assertEqual(all_bert_files["module_name"], "bert")
bert_model_patterns = bert_info["model_patterns"]
self.assertEqual(bert_model_patterns.model_name, "BERT")
self.assertEqual(bert_model_patterns.checkpoint, "google-bert/bert-base-uncased")
self.assertEqual(bert_model_patterns.model_type, "bert")
self.assertEqual(bert_model_patterns.model_lower_cased, "bert")
self.assertEqual(bert_model_patterns.model_camel_cased, "Bert")
self.assertEqual(bert_model_patterns.model_upper_cased, "BERT")
self.assertEqual(bert_model_patterns.config_class, "BertConfig")
self.assertEqual(bert_model_patterns.tokenizer_class, "BertTokenizer")
self.assertIsNone(bert_model_patterns.feature_extractor_class)
self.assertIsNone(bert_model_patterns.processor_class)
def test_retrieve_info_for_model_with_vit(self):
vit_info = retrieve_info_for_model("vit")
vit_classes = ["ViTForImageClassification", "ViTModel"]

127
tests/utils/test_modeling_tf_utils.py
View File

@@ -16,7 +16,6 @@
from __future__ import annotations
import inspect
import json
import os
import random
@@ -24,10 +23,10 @@ import tempfile
import unittest
import unittest.mock as mock
from huggingface_hub import HfFolder, Repository, snapshot_download
from huggingface_hub import HfFolder, snapshot_download
from requests.exceptions import HTTPError
from transformers import is_tf_available, is_torch_available
from transformers import is_tf_available
from transformers.configuration_utils import PretrainedConfig
from transformers.testing_utils import ( # noqa: F401
TOKEN,
@@ -35,11 +34,9 @@ from transformers.testing_utils import ( # noqa: F401
CaptureLogger,
TemporaryHubRepo,
_tf_gpu_memory_limit,
is_pt_tf_cross_test,
is_staging_test,
require_safetensors,
require_tf,
require_torch,
slow,
)
from transformers.utils import (
@@ -61,14 +58,9 @@ if is_tf_available():
from transformers import (
BertConfig,
PreTrainedModel,
PushToHubCallback,
RagRetriever,
TFAutoModel,
TFBertForMaskedLM,
TFBertForSequenceClassification,
TFBertModel,
TFPreTrainedModel,
TFRagModel,
)
from transformers.modeling_tf_utils import keras, tf_shard_checkpoint, unpack_inputs
@@ -90,9 +82,6 @@ if is_tf_available():
# Virtual devices must be set before GPUs have been initialized
print(e)
if is_torch_available():
from transformers import BertModel
@require_tf
class TFModelUtilsTest(unittest.TestCase):
@@ -241,34 +230,6 @@ class TFModelUtilsTest(unittest.TestCase):
# If this doesn't throw an error then the test passes
TFBertForSequenceClassification.from_pretrained("ArthurZ/tiny-random-bert-sharded")
@is_pt_tf_cross_test
def test_checkpoint_sharding_local_from_pt(self):
with tempfile.TemporaryDirectory() as tmp_dir:
_ = Repository(local_dir=tmp_dir, clone_from="hf-internal-testing/tiny-random-bert-sharded")
model = TFBertModel.from_pretrained(tmp_dir, from_pt=True)
# the model above is the same as the model below, just a sharded pytorch version.
ref_model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
for p1, p2 in zip(model.weights, ref_model.weights):
assert np.allclose(p1.numpy(), p2.numpy())
@is_pt_tf_cross_test
def test_checkpoint_loading_with_prefix_from_pt(self):
model = TFBertModel.from_pretrained(
"hf-internal-testing/tiny-random-bert", from_pt=True, load_weight_prefix="a/b"
)
ref_model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert", from_pt=True)
for p1, p2 in zip(model.weights, ref_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
self.assertTrue(p1.name.startswith("a/b/"))
@is_pt_tf_cross_test
def test_checkpoint_sharding_hub_from_pt(self):
model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert-sharded", from_pt=True)
# the model above is the same as the model below, just a sharded pytorch version.
ref_model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
for p1, p2 in zip(model.weights, ref_model.weights):
assert np.allclose(p1.numpy(), p2.numpy())
def test_shard_checkpoint(self):
# This is the model we will use, total size 340,000 bytes.
model = keras.Sequential(
@@ -437,16 +398,6 @@ class TFModelUtilsTest(unittest.TestCase):
for p1, p2 in zip(model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@is_pt_tf_cross_test
@require_safetensors
def test_bfloat16_torch_loading(self):
# Assert that neither of these raise an error - both repos contain bfloat16 tensors
model1 = TFAutoModel.from_pretrained("Rocketknight1/tiny-random-gpt2-bfloat16-pt", from_pt=True)
model2 = TFAutoModel.from_pretrained("Rocketknight1/tiny-random-gpt2-bfloat16") # PT-format safetensors
# Check that PT and safetensors loading paths end up with the same values
for weight1, weight2 in zip(model1.weights, model2.weights):
self.assertTrue(tf.reduce_all(weight1 == weight2))
@slow
def test_save_pretrained_signatures(self):
model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
@@ -522,36 +473,6 @@ class TFModelUtilsTest(unittest.TestCase):
for p1, p2 in zip(model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@is_pt_tf_cross_test
def test_safetensors_save_and_load_pt_to_tf(self):
model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
pt_model = BertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
with tempfile.TemporaryDirectory() as tmp_dir:
pt_model.save_pretrained(tmp_dir, safe_serialization=True)
# Check we have a model.safetensors file
self.assertTrue(os.path.isfile(os.path.join(tmp_dir, SAFE_WEIGHTS_NAME)))
new_model = TFBertModel.from_pretrained(tmp_dir)
# Check models are equal
for p1, p2 in zip(model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@is_pt_tf_cross_test
def test_sharded_safetensors_save_and_load_pt_to_tf(self):
model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
pt_model = BertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
with tempfile.TemporaryDirectory() as tmp_dir:
pt_model.save_pretrained(tmp_dir, safe_serialization=True, max_shard_size="150kB")
# Check we have a safetensors shard index file
self.assertTrue(os.path.isfile(os.path.join(tmp_dir, SAFE_WEIGHTS_INDEX_NAME)))
new_model = TFBertModel.from_pretrained(tmp_dir)
# Check models are equal
for p1, p2 in zip(model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@require_safetensors
def test_safetensors_load_from_hub(self):
tf_model = TFBertModel.from_pretrained("hf-internal-testing/tiny-random-bert")
@@ -581,19 +502,6 @@ class TFModelUtilsTest(unittest.TestCase):
for p1, p2 in zip(model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@require_safetensors
@is_pt_tf_cross_test
def test_safetensors_tf_from_torch(self):
hub_model = TFBertModel.from_pretrained("hf-internal-testing/tiny-bert-tf-only")
model = BertModel.from_pretrained("hf-internal-testing/tiny-bert-pt-only")
with tempfile.TemporaryDirectory() as tmp_dir:
model.save_pretrained(tmp_dir, safe_serialization=True)
new_model = TFBertModel.from_pretrained(tmp_dir)
for p1, p2 in zip(hub_model.weights, new_model.weights):
self.assertTrue(np.allclose(p1.numpy(), p2.numpy()))
@require_safetensors
def test_safetensors_tf_from_sharded_h5_with_sharded_safetensors_local(self):
with tempfile.TemporaryDirectory() as tmp_dir:
@@ -729,37 +637,6 @@ class TFModelPushToHubTester(unittest.TestCase):
break
self.assertTrue(models_equal)
@is_pt_tf_cross_test
def test_push_to_hub_callback(self):
with TemporaryHubRepo(token=self._token) as tmp_repo:
config = BertConfig(
vocab_size=99, hidden_size=32, num_hidden_layers=5, num_attention_heads=4, intermediate_size=37
)
model = TFBertForMaskedLM(config)
model.compile()
with tempfile.TemporaryDirectory() as tmp_dir:
push_to_hub_callback = PushToHubCallback(
output_dir=tmp_dir,
hub_model_id=tmp_repo.repo_id,
hub_token=self._token,
)
model.fit(model.dummy_inputs, model.dummy_inputs, epochs=1, callbacks=[push_to_hub_callback])
new_model = TFBertForMaskedLM.from_pretrained(tmp_repo.repo_id)
models_equal = True
for p1, p2 in zip(model.weights, new_model.weights):
if not tf.math.reduce_all(p1 == p2):
models_equal = False
break
self.assertTrue(models_equal)
tf_push_to_hub_params = dict(inspect.signature(TFPreTrainedModel.push_to_hub).parameters)
tf_push_to_hub_params.pop("base_model_card_args")
pt_push_to_hub_params = dict(inspect.signature(PreTrainedModel.push_to_hub).parameters)
pt_push_to_hub_params.pop("deprecated_kwargs")
self.assertDictEaual(tf_push_to_hub_params, pt_push_to_hub_params)
def test_push_to_hub_in_organization(self):
with TemporaryHubRepo(namespace="valid_org", token=self._token) as tmp_repo:
config = BertConfig(