From d051212eb3435491b093b74d7dceab3c634b7101 Mon Sep 17 00:00:00 2001
From: shengyang wang <shengyang.wang@bytedance.com>
Date: Sat, 9 Nov 2024 15:12:19 -0800
Subject: [PATCH] Update Dockerfile with CUDA 11.8
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

docker image里的cuda好像要升级一下，不然有这个报错: server log:2024-11-09 12:31:36 2024-11-09 20:31:36,085 INFO send inference response
2024-11-09 12:31:36
  0%|          | 0/1 [00:00<?, ?it/s]2024-11-09 20:31:36,093 INFO synthesis text 你好,我是通义生成式语音大模型,请问有什么可以帮您的吗?
2024-11-09 12:31:44
  0%|          | 0/1 [00:08<?, ?it/s]
2024-11-09 12:31:44 2024-11-09 20:31:44,105 ERROR Exception iterating responses: cuFFT error: CUFFT_INTERNAL_ERROR
2024-11-09 12:31:44 Traceback (most recent call last):
2024-11-09 12:31:44   File "/opt/conda/lib/python3.10/site-packages/grpc/_server.py", line 589, in _take_response_from_response_iterator
2024-11-09 12:31:44     return next(response_iterator), True
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/server.py", line 62, in Inference
2024-11-09 12:31:44     for i in model_output:
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/cli/cosyvoice.py", line 61, in inference_sft
2024-11-09 12:31:44     for model_output in self.model.tts(**model_input, stream=stream, speed=speed):
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/cli/model.py", line 191, in tts
2024-11-09 12:31:44     this_tts_speech = self.token2wav(token=this_tts_speech_token,
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/cli/model.py", line 138, in token2wav
2024-11-09 12:31:44     tts_speech, tts_source = self.hift.inference(speech_feat=tts_mel, cache_source=hift_cache_source)
2024-11-09 12:31:44   File "/opt/conda/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
2024-11-09 12:31:44     return func(*args, **kwargs)
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/hifigan/generator.py", line 410, in inference
2024-11-09 12:31:44     generated_speech = self.decode(x=speech_feat, s=s)
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/hifigan/generator.py", line 350, in decode
2024-11-09 12:31:44     s_stft_real, s_stft_imag = self._stft(s.squeeze(1))
2024-11-09 12:31:44   File "/opt/CosyVoice/CosyVoice/runtime/python/grpc/../../../cosyvoice/hifigan/generator.py", line 334, in _stft
2024-11-09 12:31:44     spec = torch.stft(
2024-11-09 12:31:44   File "/opt/conda/lib/python3.10/site-packages/torch/functional.py", line 641, in stft
2024-11-09 12:31:44     return _VF.stft(input, n_fft, hop_length, win_length, window,  # type: ignore[attr-defined]
2024-11-09 12:31:44 RuntimeError: cuFFT error: CUFFT_INTERNAL_ERROR
---
 runtime/python/Dockerfile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/runtime/python/Dockerfile b/runtime/python/Dockerfile
index 6b5e9442..6f35ce5b 100644
--- a/runtime/python/Dockerfile
+++ b/runtime/python/Dockerfile
@@ -1,4 +1,4 @@
-FROM pytorch/pytorch:2.0.1-cuda11.7-cudnn8-runtime
+FROM pytorch/pytorch:2.5.1-cuda11.8-cudnn9-runtime
 ENV DEBIAN_FRONTEND=noninteractive
 
 WORKDIR /opt/CosyVoice
@@ -10,4 +10,4 @@ RUN git lfs install
 RUN git clone --recursive https://github.com/FunAudioLLM/CosyVoice.git
 # here we use python==3.10 because we cannot find an image which have both python3.8 and torch2.0.1-cu118 installed
 RUN cd CosyVoice && pip3 install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host=mirrors.aliyun.com
-RUN cd CosyVoice/runtime/python/grpc && python3 -m grpc_tools.protoc -I. --python_out=. --grpc_python_out=. cosyvoice.proto
\ No newline at end of file
+RUN cd CosyVoice/runtime/python/grpc && python3 -m grpc_tools.protoc -I. --python_out=. --grpc_python_out=. cosyvoice.proto