Commit @4468033129925a87f3332f45002b1bacc328a918 - yjyoon/whisper_server

Fedir Zadniprovskyi 2024-07-03

feat: add language field to the model response type

@4468033129925a87f3332f45002b1bacc328a918

b472650

4468033

faster_whisper_server/main.py

--- faster_whisper_server/main.py

+++ faster_whisper_server/main.py


 
 @app.get("/v1/models")
 def get_models() -> ModelListResponse:
-    models = huggingface_hub.list_models(library="ctranslate2", tags="automatic-speech-recognition")
-    models = [
-        ModelObject(
+    models = huggingface_hub.list_models(library="ctranslate2", tags="automatic-speech-recognition", cardData=True)
+    transformed_models: list[ModelObject] = []
+    for model in models:
+        assert model.created_at is not None
+        assert model.card_data is not None
+        assert model.card_data.language is None or isinstance(model.card_data.language, str | list)
+        if model.card_data.language is None:
+            language = []
+        elif isinstance(model.card_data.language, str):
+            language = [model.card_data.language]
+        else:
+            language = model.card_data.language
+        transformed_model = ModelObject(
             id=model.id,
             created=int(model.created_at.timestamp()),
             object_="model",
             owned_by=model.id.split("/")[0],
+            language=language,
         )
-        for model in models
-        if model.created_at is not None
-    ]
-    return ModelListResponse(data=models)
+        transformed_models.append(transformed_model)
+    return ModelListResponse(data=transformed_models)
 
 
 @app.get("/v1/models/{model_name:path}")

     model_name: Annotated[str, Path(example="Systran/faster-distil-whisper-large-v3")],
 ) -> ModelObject:
     models = list(
-        huggingface_hub.list_models(model_name=model_name, library="ctranslate2", tags="automatic-speech-recognition")
+        huggingface_hub.list_models(
+            model_name=model_name, library="ctranslate2", tags="automatic-speech-recognition", cardData=True
+        )
     )
     if len(models) == 0:
         raise HTTPException(status_code=404, detail="Model doesn't exists")

             detail=f"Model doesn't exists. Possible matches: {", ".join([model.id for model in models])}",
         )
     assert exact_match.created_at is not None
+    assert exact_match.card_data is not None
+    assert exact_match.card_data.language is None or isinstance(exact_match.card_data.language, str | list)
+    if exact_match.card_data.language is None:
+        language = []
+    elif isinstance(exact_match.card_data.language, str):
+        language = [exact_match.card_data.language]
+    else:
+        language = exact_match.card_data.language
     return ModelObject(
         id=exact_match.id,
         created=int(exact_match.created_at.timestamp()),
         object_="model",
         owned_by=exact_match.id.split("/")[0],
+        language=language,
     )
 
 

b472650

4468033

faster_whisper_server/server_models.py

--- faster_whisper_server/server_models.py

+++ faster_whisper_server/server_models.py


     """The object type, which is always "model"."""
     owned_by: str
     """The organization that owns the model."""
+    language: list[str] = Field(default_factory=list)
+    """List of ISO 639-3 supported by the model. It's possible that the list will be empty. This field is not a part of the OpenAI API spec and is added for convenience."""  # noqa: E501
 
     model_config = ConfigDict(
         populate_by_name=True,

Add a comment

Open 0
Closed 0

List

...	...	@@ -87,18 +87,27 @@
87	87
88	88	@app.get("/v1/models")
89	89	def get_models() -> ModelListResponse:
90		- models = huggingface_hub.list_models(library="ctranslate2", tags="automatic-speech-recognition")
91		- models = [
92		- ModelObject(
	90	+ models = huggingface_hub.list_models(library="ctranslate2", tags="automatic-speech-recognition", cardData=True)
	91	+ transformed_models: list[ModelObject] = []
	92	+ for model in models:
	93	+ assert model.created_at is not None
	94	+ assert model.card_data is not None
	95	+ assert model.card_data.language is None or isinstance(model.card_data.language, str \| list)
	96	+ if model.card_data.language is None:
	97	+ language = []
	98	+ elif isinstance(model.card_data.language, str):
	99	+ language = [model.card_data.language]
	100	+ else:
	101	+ language = model.card_data.language
	102	+ transformed_model = ModelObject(
93	103	id=model.id,
94	104	created=int(model.created_at.timestamp()),
95	105	object_="model",
96	106	owned_by=model.id.split("/")[0],
	107	+ language=language,
97	108	)
98		- for model in models
99		- if model.created_at is not None
100		- ]
101		- return ModelListResponse(data=models)
	109	+ transformed_models.append(transformed_model)
	110	+ return ModelListResponse(data=transformed_models)
102	111
103	112
104	113	@app.get("/v1/models/{model_name:path}")
...	...	@@ -107,7 +116,9 @@
107	116	model_name: Annotated[str, Path(example="Systran/faster-distil-whisper-large-v3")],
108	117	) -> ModelObject:
109	118	models = list(
110		- huggingface_hub.list_models(model_name=model_name, library="ctranslate2", tags="automatic-speech-recognition")
	119	+ huggingface_hub.list_models(
	120	+ model_name=model_name, library="ctranslate2", tags="automatic-speech-recognition", cardData=True
	121	+ )
111	122	)
112	123	if len(models) == 0:
113	124	raise HTTPException(status_code=404, detail="Model doesn't exists")
...	...	@@ -122,11 +133,20 @@
122	133	detail=f"Model doesn't exists. Possible matches: {", ".join([model.id for model in models])}",
123	134	)
124	135	assert exact_match.created_at is not None
	136	+ assert exact_match.card_data is not None
	137	+ assert exact_match.card_data.language is None or isinstance(exact_match.card_data.language, str \| list)
	138	+ if exact_match.card_data.language is None:
	139	+ language = []
	140	+ elif isinstance(exact_match.card_data.language, str):
	141	+ language = [exact_match.card_data.language]
	142	+ else:
	143	+ language = exact_match.card_data.language
125	144	return ModelObject(
126	145	id=exact_match.id,
127	146	created=int(exact_match.created_at.timestamp()),
128	147	object_="model",
129	148	owned_by=exact_match.id.split("/")[0],
	149	+ language=language,
130	150	)
131	151
132	152

...	...	@@ -133,6 +133,8 @@
133	133	"""The object type, which is always "model"."""
134	134	owned_by: str
135	135	"""The organization that owns the model."""
	136	+ language: list[str] = Field(default_factory=list)
	137	+ """List of ISO 639-3 supported by the model. It's possible that the list will be empty. This field is not a part of the OpenAI API spec and is added for convenience.""" # noqa: E501
136	138
137	139	model_config = ConfigDict(
138	140	populate_by_name=True,

Delete comment