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

Chat template: save and load correctly for processors (#33462)

Browse Source 
* fix

* add tests

* fix tests

* Update tests/models/llava/test_processor_llava.py

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* fix

* fix tests

* update tests

---------

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
This commit is contained in:
Raushan Turganbay
2024-09-18 13:00:44 +02:00
committed by GitHub
parent 52e22cbf67
commit db72894b48
4 changed files with 106 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
tests/models/llava_onevision/test_processing_llava_onevision.py
View File

@@ -11,6 +11,7 @@
# 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 json
import shutil
import tempfile
import unittest
@@ -40,9 +41,10 @@ class LlavaOnevisionProcessorTest(ProcessorTesterMixin, unittest.TestCase):
image_processor = LlavaOnevisionImageProcessor()
video_processor = LlavaOnevisionVideoProcessor()
tokenizer = Qwen2TokenizerFast.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
processor_kwargs = self.prepare_processor_dict()
processor = LlavaOnevisionProcessor(
video_processor=video_processor, image_processor=image_processor, tokenizer=tokenizer
video_processor=video_processor, image_processor=image_processor, tokenizer=tokenizer, **processor_kwargs
)
processor.save_pretrained(self.tmpdirname)
@@ -52,9 +54,32 @@ class LlavaOnevisionProcessorTest(ProcessorTesterMixin, unittest.TestCase):
def get_image_processor(self, **kwargs):
return AutoProcessor.from_pretrained(self.tmpdirname, **kwargs).image_processor
def get_Video_processor(self, **kwargs):
def get_video_processor(self, **kwargs):
return AutoProcessor.from_pretrained(self.tmpdirname, **kwargs).video_processor
def prepare_processor_dict(self):
return {"chat_template": "dummy_template"}
@unittest.skip(
"Skip because the model has no processor kwargs except for chat template and"
"chat template is saved as a separate file. Stop skipping this test when the processor"
"has new kwargs saved in config file."
)
def test_processor_to_json_string(self):
pass
# Copied from tests.models.llava.test_processor_llava.LlavaProcessorTest.test_chat_template_is_saved
def test_chat_template_is_saved(self):
processor_loaded = self.processor_class.from_pretrained(self.tmpdirname)
processor_dict_loaded = json.loads(processor_loaded.to_json_string())
# chat templates aren't serialized to json in processors
self.assertFalse("chat_template" in processor_dict_loaded.keys())
# they have to be saved as separate file and loaded back from that file
# so we check if the same template is loaded
processor_dict = self.prepare_processor_dict()
self.assertTrue(processor_loaded.chat_template == processor_dict.get("chat_template", None))
def tearDown(self):
shutil.rmtree(self.tmpdirname)