Commit @3580f89aacb57c8e4c5ce2abae35cc08bf4ac766 - yjyoon/whisper_server

Fedir Zadniprovskyi 01-13

fix: tests

@3580f89aacb57c8e4c5ce2abae35cc08bf4ac766

1bb8819

3580f89

.github/workflows/lint.yaml

--- .github/workflows/lint.yaml

+++ .github/workflows/lint.yaml


       - uses: actions/checkout@v4
       - uses: astral-sh/setup-uv@v5
         with:
-          version: "0.4.11"
+          version: "latest"
           enable-cache: true
       - run: uv python install 3.12
       - run: uv sync --extra dev

1bb8819

3580f89

.github/workflows/publish-docs.yaml

--- .github/workflows/publish-docs.yaml

+++ .github/workflows/publish-docs.yaml


       - uses: actions/checkout@v4
       - uses: astral-sh/setup-uv@v5
         with:
-          version: "0.4.11"
+          version: "latest"
           enable-cache: true
       - run: uv python install 3.12
       - run: uv sync --extra dev

1bb8819

3580f89

.github/workflows/test.yaml

--- .github/workflows/test.yaml

+++ .github/workflows/test.yaml


       - uses: actions/checkout@v4
       - uses: astral-sh/setup-uv@v5
         with:
-          version: "0.4.11"
+          version: "latest"
           enable-cache: true
       - run: uv python install 3.12
       - run: uv sync --all-extras

1bb8819

3580f89

src/speaches/hf_utils.py

--- src/speaches/hf_utils.py

+++ src/speaches/hf_utils.py


 def download_kokoro_model() -> None:
     model_id = "hexgrad/Kokoro-82M"
     model_repo_path = Path(
-        huggingface_hub.snapshot_download(model_id, repo_type="model", allow_patterns="**/kokoro-v0_19.onnx")
+        huggingface_hub.snapshot_download(model_id, repo_type="model", allow_patterns=["kokoro-v0_19.onnx"])
     )
     # HACK
     res = httpx.get(
         "https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files/voices.json", follow_redirects=True
     ).raise_for_status()
     voices_path = model_repo_path / "voices.json"
+    voices_path.touch(exist_ok=True)
     voices_path.write_bytes(res.content)
 
 

1bb8819

3580f89

tests/conftest.py

--- tests/conftest.py

+++ tests/conftest.py


 from typing import Protocol
 
 from fastapi.testclient import TestClient
+import httpx
 from httpx import ASGITransport, AsyncClient
-from huggingface_hub import snapshot_download
 from openai import AsyncOpenAI
 import pytest
 import pytest_asyncio

 
 from speaches.config import Config, WhisperConfig
 from speaches.dependencies import get_config
+from speaches.hf_utils import download_kokoro_model
 from speaches.main import create_app
 
 DISABLE_LOGGERS = ["multipart.multipart", "faster_whisper"]

     # disable the UI as it slightly increases the app startup time due to the imports it's doing
     enable_ui=False,
 )
+TIMEOUT = httpx.Timeout(15.0)
 
 
 def pytest_configure() -> None:

         app = create_app()
         # https://fastapi.tiangolo.com/advanced/testing-dependencies/
         app.dependency_overrides[get_config] = lambda: config
-        async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as aclient:
+        async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test", timeout=TIMEOUT) as aclient:
             yield aclient
 
     return inner

 
 # TODO: remove the download after running the tests
 # TODO: do not download when not needed
+# @pytest.fixture(scope="session", autouse=True)
+# def download_piper_voices() -> None:
+#     # Only download `voices.json` and the default voice
+#     snapshot_download("rhasspy/piper-voices", allow_patterns=["voices.json", "en/en_US/amy/**"])
+
+
 @pytest.fixture(scope="session", autouse=True)
-def download_piper_voices() -> None:
-    # Only download `voices.json` and the default voice
-    snapshot_download("rhasspy/piper-voices", allow_patterns=["voices.json", "en/en_US/amy/**"])
+def download_kokoro() -> None:
+    download_kokoro_model()

1bb8819

3580f89

tests/speech_test.py

--- tests/speech_test.py

+++ tests/speech_test.py


 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize("response_format", SUPPORTED_RESPONSE_FORMATS)
 async def test_create_speech_formats(openai_client: AsyncOpenAI, response_format: ResponseFormat) -> None:
     await openai_client.audio.speech.create(

 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize(("model", "voice"), GOOD_MODEL_VOICE_PAIRS)
 async def test_create_speech_good_model_voice_pair(openai_client: AsyncOpenAI, model: str, voice: str) -> None:
     await openai_client.audio.speech.create(

 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize(("model", "voice"), BAD_MODEL_VOICE_PAIRS)
 async def test_create_speech_bad_model_voice_pair(openai_client: AsyncOpenAI, model: str, voice: str) -> None:
     # NOTE: not sure why `APIConnectionError` is sometimes raised

         )
 
 
-SUPPORTED_SPEEDS = [0.25, 0.5, 1.0, 2.0, 4.0]
+SUPPORTED_SPEEDS = [0.5, 1.0, 2.0]
 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 async def test_create_speech_with_varying_speed(openai_client: AsyncOpenAI) -> None:
     previous_size: int | None = None
     for speed in SUPPORTED_SPEEDS:

 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize("speed", UNSUPPORTED_SPEEDS)
 async def test_create_speech_with_unsupported_speed(openai_client: AsyncOpenAI, speed: float) -> None:
     with pytest.raises(UnprocessableEntityError):

 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize("sample_rate", VALID_SAMPLE_RATES)
 async def test_speech_valid_resample(openai_client: AsyncOpenAI, sample_rate: int) -> None:
     res = await openai_client.audio.speech.create(

 
 
 @pytest.mark.asyncio
-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
 @pytest.mark.parametrize("sample_rate", INVALID_SAMPLE_RATES)
 async def test_speech_invalid_resample(openai_client: AsyncOpenAI, sample_rate: int) -> None:
     with pytest.raises(UnprocessableEntityError):

         )
 
 
+# TODO: add piper tests
+
 # TODO: implement the following test
 
 # NUMBER_OF_MODELS = 1

Add a comment

Open 0
Closed 0

List

...	...	@@ -22,7 +22,7 @@
22	22	- uses: actions/checkout@v4
23	23	- uses: astral-sh/setup-uv@v5
24	24	with:
25		- version: "0.4.11"
	25	+ version: "latest"
26	26	enable-cache: true
27	27	- run: uv python install 3.12
28	28	- run: uv sync --extra dev

...	...	@@ -20,7 +20,7 @@
20	20	- uses: actions/checkout@v4
21	21	- uses: astral-sh/setup-uv@v5
22	22	with:
23		- version: "0.4.11"
	23	+ version: "latest"
24	24	enable-cache: true
25	25	- run: uv python install 3.12
26	26	- run: uv sync --extra dev

...	...	@@ -24,7 +24,7 @@
24	24	- uses: actions/checkout@v4
25	25	- uses: astral-sh/setup-uv@v5
26	26	with:
27		- version: "0.4.11"
	27	+ version: "latest"
28	28	enable-cache: true
29	29	- run: uv python install 3.12
30	30	- run: uv sync --all-extras

...	...	@@ -226,13 +226,14 @@
226	226	def download_kokoro_model() -> None:
227	227	model_id = "hexgrad/Kokoro-82M"
228	228	model_repo_path = Path(
229		- huggingface_hub.snapshot_download(model_id, repo_type="model", allow_patterns="**/kokoro-v0_19.onnx")
	229	+ huggingface_hub.snapshot_download(model_id, repo_type="model", allow_patterns=["kokoro-v0_19.onnx"])
230	230	)
231	231	# HACK
232	232	res = httpx.get(
233	233	"https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files/voices.json", follow_redirects=True
234	234	).raise_for_status()
235	235	voices_path = model_repo_path / "voices.json"
	236	+ voices_path.touch(exist_ok=True)
236	237	voices_path.write_bytes(res.content)
237	238
238	239

...	...	@@ -5,8 +5,8 @@
5	5	from typing import Protocol
6	6
7	7	from fastapi.testclient import TestClient
	8	+import httpx
8	9	from httpx import ASGITransport, AsyncClient
9		-from huggingface_hub import snapshot_download
10	10	from openai import AsyncOpenAI
11	11	import pytest
12	12	import pytest_asyncio
...	...	@@ -14,6 +14,7 @@
14	14
15	15	from speaches.config import Config, WhisperConfig
16	16	from speaches.dependencies import get_config
	17	+from speaches.hf_utils import download_kokoro_model
17	18	from speaches.main import create_app
18	19
19	20	DISABLE_LOGGERS = ["multipart.multipart", "faster_whisper"]
...	...	@@ -26,6 +27,7 @@
26	27	# disable the UI as it slightly increases the app startup time due to the imports it's doing
27	28	enable_ui=False,
28	29	)
	30	+TIMEOUT = httpx.Timeout(15.0)
29	31
30	32
31	33	def pytest_configure() -> None:
...	...	@@ -64,7 +66,7 @@
64	66	app = create_app()
65	67	# https://fastapi.tiangolo.com/advanced/testing-dependencies/
66	68	app.dependency_overrides[get_config] = lambda: config
67		- async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as aclient:
	69	+ async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test", timeout=TIMEOUT) as aclient:
68	70	yield aclient
69	71
70	72	return inner
...	...	@@ -91,7 +93,12 @@
91	93
92	94	# TODO: remove the download after running the tests
93	95	# TODO: do not download when not needed
	96	+# @pytest.fixture(scope="session", autouse=True)
	97	+# def download_piper_voices() -> None:
	98	+# # Only download `voices.json` and the default voice
	99	+# snapshot_download("rhasspy/piper-voices", allow_patterns=["voices.json", "en/en_US/amy/**"])
	100	+
	101	+
94	102	@pytest.fixture(scope="session", autouse=True)
95		-def download_piper_voices() -> None:
96		- # Only download `voices.json` and the default voice
97		- snapshot_download("rhasspy/piper-voices", allow_patterns=["voices.json", "en/en_US/amy/**"])
	103	+def download_kokoro() -> None:
	104	+ download_kokoro_model()

...	...	@@ -21,7 +21,6 @@
21	21
22	22
23	23	@pytest.mark.asyncio
24		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
25	24	@pytest.mark.parametrize("response_format", SUPPORTED_RESPONSE_FORMATS)
26	25	async def test_create_speech_formats(openai_client: AsyncOpenAI, response_format: ResponseFormat) -> None:
27	26	await openai_client.audio.speech.create(
...	...	@@ -42,7 +41,6 @@
42	41
43	42
44	43	@pytest.mark.asyncio
45		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
46	44	@pytest.mark.parametrize(("model", "voice"), GOOD_MODEL_VOICE_PAIRS)
47	45	async def test_create_speech_good_model_voice_pair(openai_client: AsyncOpenAI, model: str, voice: str) -> None:
48	46	await openai_client.audio.speech.create(
...	...	@@ -63,7 +61,6 @@
63	61
64	62
65	63	@pytest.mark.asyncio
66		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
67	64	@pytest.mark.parametrize(("model", "voice"), BAD_MODEL_VOICE_PAIRS)
68	65	async def test_create_speech_bad_model_voice_pair(openai_client: AsyncOpenAI, model: str, voice: str) -> None:
69	66	# NOTE: not sure why `APIConnectionError` is sometimes raised
...	...	@@ -76,11 +73,10 @@
76	73	)
77	74
78	75
79		-SUPPORTED_SPEEDS = [0.25, 0.5, 1.0, 2.0, 4.0]
	76	+SUPPORTED_SPEEDS = [0.5, 1.0, 2.0]
80	77
81	78
82	79	@pytest.mark.asyncio
83		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
84	80	async def test_create_speech_with_varying_speed(openai_client: AsyncOpenAI) -> None:
85	81	previous_size: int \| None = None
86	82	for speed in SUPPORTED_SPEEDS:
...	...	@@ -101,7 +97,6 @@
101	97
102	98
103	99	@pytest.mark.asyncio
104		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
105	100	@pytest.mark.parametrize("speed", UNSUPPORTED_SPEEDS)
106	101	async def test_create_speech_with_unsupported_speed(openai_client: AsyncOpenAI, speed: float) -> None:
107	102	with pytest.raises(UnprocessableEntityError):
...	...	@@ -118,7 +113,6 @@
118	113
119	114
120	115	@pytest.mark.asyncio
121		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
122	116	@pytest.mark.parametrize("sample_rate", VALID_SAMPLE_RATES)
123	117	async def test_speech_valid_resample(openai_client: AsyncOpenAI, sample_rate: int) -> None:
124	118	res = await openai_client.audio.speech.create(
...	...	@@ -136,7 +130,6 @@
136	130
137	131
138	132	@pytest.mark.asyncio
139		-@pytest.mark.skipif(platform_machine != "x86_64", reason="Only supported on x86_64")
140	133	@pytest.mark.parametrize("sample_rate", INVALID_SAMPLE_RATES)
141	134	async def test_speech_invalid_resample(openai_client: AsyncOpenAI, sample_rate: int) -> None:
142	135	with pytest.raises(UnprocessableEntityError):
...	...	@@ -149,6 +142,8 @@
149	142	)
150	143
151	144
	145	+# TODO: add piper tests
	146	+
152	147	# TODO: implement the following test
153	148
154	149	# NUMBER_OF_MODELS = 1

Delete comment