0.1B INT8 的 AR token 生成链路异常
INFO: 127.0.0.1:55774 - "POST /api/tts HTTP/1.1" 500 Internal Server Error
ERROR: Exception in ASGI application
Traceback (most recent call last):
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/uvicorn/protocols/http/h11_impl.py", line 416, in run_asgi
result = await app( # type: ignore[func-returns-value]
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/uvicorn/middleware/proxy_headers.py", line 63, in __call__
return await self.app(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/applications.py", line 1163, in __call__
await super().__call__(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/applications.py", line 96, in __call__
await self.middleware_stack(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/middleware/errors.py", line 186, in __call__
raise exc
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/middleware/errors.py", line 164, in __call__
await self.app(scope, receive, _send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/middleware/exceptions.py", line 63, in __call__
await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
raise exc
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
await app(scope, receive, sender)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/middleware/asyncexitstack.py", line 18, in __call__
await self.app(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/routing.py", line 670, in __call__
await self.middleware_stack(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 2734, in app
await route.handle(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 1281, in handle
await super().handle(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/routing.py", line 280, in handle
await self.app(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 158, in app
await wrap_app_handling_exceptions(app, request)(scope, receive, send)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
raise exc
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
await app(scope, receive, sender)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 144, in app
response = await f(request)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 706, in app
raw_response = await run_endpoint_function(
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/fastapi/routing.py", line 354, in run_endpoint_function
return await run_in_threadpool(dependant.call, **values)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/starlette/concurrency.py", line 34, in run_in_threadpool
return await anyio.to_thread.run_sync(func)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/anyio/to_thread.py", line 65, in run_sync
return await get_async_backend().run_sync_in_worker_thread(
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 2641, in run_sync_in_worker_thread
return await future
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/.venv/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 1033, in run
result = context.run(func, *args)
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/onnx_runtime_0_1b_int8/arktts_runtime/service.py", line 202, in tts
audio, _ = obj.synthesize(
File "/home/yyl/RapidSpeech.cpp/Audio8_TTS/onnx_runtime_0_1b_int8/arktts_runtime/runtime.py", line 436, in synthesize
raise RuntimeError("model produced no codec frames")
RuntimeError: model produced no codec frames
onnx 0.1b int8生成tts失败，0.6b的可以成功，如果根据readme来，注册的default音色可以顺利推理，但是音频是乱码，不可用。
感谢反馈。我们已定位并修复 /api/tts 返回 500 的问题。
- GitHub commit: https://github.com/Audio8-AI/Audio8_TTS/commit/6bdc912
- 修改目录：
onnx_runtime_0_1b_int8/
原因
0.1B Slow AR 输出 4097 维 compact logits，其中最后一项是 EOS。原运行时在首个生成步采样到 EOS 时会直接结束，未生成任何 codec frame，最终触发：
RuntimeError: model produced no codec frames
INT8 量化和不同 ORT CPU kernel 可能导致首步 EOS 在不同环境下偶发出现。
请使用最新的 GitHub master 和该 HF 模型版本测试。如仍有问题，请附上请求 JSON、操作系统、CPU、onnxruntime 版本、seed，以及生成的 .npy 或 WAV 信息。
------------ English Version --------------
Thanks for reporting this. We tracked down and fixed the /api/tts HTTP 500 issue.
- GitHub commit: https://github.com/Audio8-AI/Audio8_TTS/commit/6bdc912
- Changed directory:
onnx_runtime_0_1b_int8/
The 0.1B Slow AR model outputs 4,097 compact logits: 4,096 semantic logits plus one EOS logit. With the old runtime, if EOS was sampled on the very first generation step, decoding stopped immediately without producing any codec frames. That eventually caused "RuntimeError: model produced no codec frames".
Because of INT8 quantization and small differences between ORT CPU kernels, this could happen intermittently depending on the environment and random seed. Please try the latest GitHub master branch together with this HF model revision. If the problem still occurs, please share the request JSON, OS, CPU, onnxruntime version, seed, and basic information about the generated .npy or WAV file.