Docker部署推理时间巨慢问题

[wuxx0620]

Downloading [config/path.yaml]: 100%|██████████| 309/309 [00:00<00:00, 981B/s]8.0MB/s]
Downloading [README.md]: 100%|██████████| 1.88k/1.88k [00:00<00:00, 5.70kB/s]30.8MB/s]B/s]
Downloading [asset/DVAE.safetensors]: 100%|██████████| 57.6M/57.6M [00:02<00:00, 28.5MB/s]/s]
Downloading [asset/tokenizer/special_tokens_map.json]: 100%|██████████| 7.66k/7.66k [00:00<00:00, 24.8kB/s]
Downloading [asset/DVAE_full.pt]: 100%|██████████| 57.6M/57.6M [00:02<00:00, 28.6MB/s]B/s]/s]
Downloading [asset/spk_stat.pt]: 100%|██████████| 4.16k/4.16k [00:00<00:00, 14.1kB/s]]B/s]/s]
Downloading [asset/tokenizer/tokenizer.json]: 100%|██████████| 438k/438k [00:00<00:00, 940kB/s]
Downloading [asset/tokenizer/tokenizer_config.json]: 100%|██████████| 10.8k/10.8k [00:00<00:00, 34.1kB/s]
Downloading [asset/tokenizer.pt]: 100%|██████████| 329k/329k [00:00<00:00, 655kB/s].6MB/s]/s]?B/s]
Downloading [config/vocos.yaml]: 100%|██████████| 460/460 [00:00<00:00, 1.46kB/s]s].1MB/s]/s]:00, 24.8kB/s]
Downloading [asset/Decoder.safetensors]: 100%|██████████| 98.9M/98.9M [00:03<00:00, 29.8MB/s]]
Downloading [asset/Vocos.pt]: 100%|██████████| 51.8M/51.8M [00:01<00:00, 33.8MB/s]9.0MB/s]/s]]
Downloading [asset/Vocos.safetensors]: 100%|██████████| 51.8M/51.8M [00:01<00:00, 31.6MB/s]s]s]
Downloading [asset/Embed.safetensors]: 100%|██████████| 139M/139M [00:04<00:00, 33.1MB/s] /s]]
Downloading [asset/Decoder.pt]: 100%|██████████| 98.9M/98.9M [00:05<00:00, 20.3MB/s]MB/s]B/s]/s]
Downloading [asset/GPT.pt]: 100%|██████████| 859M/859M [00:10<00:00, 85.4MB/s], 41.9MB/s]MB/s], 34.2kB/s]
Downloading [asset/gpt/model.safetensors]: 100%|██████████| 814M/814M [00:16<00:00, 51.6MB/s]]
Fetching 23 files: 100%|██████████| 23/23 [00:19<00:00, 1.18it/s]5<00:00, 17.6MB/s] 45.3MB/s]
2024-12-24 09:48:35,835 - modelscope - INFO - Download model 'AI-ModelScope/ChatTTS' successfully.
2024-12-24 09:48:35,856 - INFO - 当前设备类型:cuda859M [00:04<00:03, 98.7MB/s]0:14, 48.7MB/s]
2024-12-24 09:48:35,856 - INFO - try to load from local: /app/asset/AI-ModelScope/ChatTTSB/s]
2024-12-24 09:48:35,856 - INFO - checking assets... | 153M/814M [00:04<00:14, 49.0MB/s]
2024-12-24 09:48:36,938 - INFO - all assets are already latest.M/814M [00:09<00:06, 60.8MB/s]
2024-12-24 09:48:37,357 - INFO - vocos loaded.%|█████████▉| 813M/814M [00:16<00:00, 28.7MB/s]
2024-12-24 09:48:37,559 - INFO - dvae loaded.
2024-12-24 09:48:41,170 - WARNING - use default LlamaModel for importing TELlamaModel error: Command 'ldconfig -p | grep 'libnvrtc'' returned non-zero exit status 1.
2024-12-24 09:48:42,056 - INFO - gpt loaded.
2024-12-24 09:48:42,381 - INFO - decoder loaded.
2024-12-24 09:48:42,396 - INFO - tokenizer loaded.
2024-12-24 09:48:42,396 - INFO - all models has been initialized.
INFO: Started server process [1]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
2024-12-24 09:50:11,078 - INFO - input:text='12344' custom_voice=1983 voice='2222' temperature=0.3 top_p=0.7 top_k=20 skip_refine=0 speed=4 text_seed=42 refine_max_new_token=384 infer_max_new_token=2048 is_stream=0 prompt='' audio_format='wav'
2024-12-24 09:50:11,140 - INFO - all models has been initialized.
text: 2%|▏ | 9/384(max) [01:50, 12.23s/it]
code: 2%|▏ | 51/2048(max) [00:35, 1.43it/s]
/home/venv/lib/python3.9/site-packages/torch/nn/modules/conv.py:306: UserWarning: Plan failed with a cudnnException: CUDNN_BACKEND_EXECUTION_PLAN_DESCRIPTOR: cudnnFinalize Descriptor Failed cudnn_status: CUDNN_STATUS_NOT_SUPPORTED (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:919.)
return F.conv1d(input, weight, bias, self.stride,

[wuxx0620]

宿主机驱动版本：
NVIDIA-SMI 535.171.04 Driver Version: 535.171.04 CUDA Version: 12.2