jianchang512 / ChatTTS-ui

一个简单的本地网页界面,使用ChatTTS将文字合成为语音,同时支持对外提供API接口。A simple native web interface that uses ChatTTS to synthesize text into speech, along with support for external API interfaces.
https://pyvideotrans.com
Other
6.25k stars 738 forks source link

Windows预打包版 -生成速度非常慢 gpu30% #37

Open openainext opened 5 months ago

openainext commented 5 months ago

Windows预打包版 -生成速度非常慢 gpu30%

image
openainext commented 5 months ago

显卡4080s

jianchang512 commented 5 months ago

下载0.5升级包补丁覆盖试试

cfso1234 commented 5 months ago

0.6版本问题依然存在

iamfoolberg commented 5 months ago

我也报告一下,2080ti GPU,生成速度慢的感人。。。 ···· root@fc5fa711b680:/host/chatTTS-ui# python3 app.py 2024-06-02 00:34:11,938 - modelscope - INFO - PyTorch version 2.3.0+cu118 Found. 2024-06-02 00:34:11,939 - modelscope - INFO - Loading ast index from /root/.cache/modelscope/ast_indexer 2024-06-02 00:34:11,939 - modelscope - INFO - No valid ast index found from /root/.cache/modelscope/ast_indexer, generating ast index from prebuilt! 2024-06-02 00:34:11,994 - modelscope - INFO - Loading done! Current index file version is 1.14.0, with md5 f4f84aaa9a1673e54fbfa12e743e64c7 and a total number of 976 components indexed Downloading: 100%|████████████████████████████████████████████| 47.0/47.0 [00:00<00:00, 304kB/s] Downloading: 100%|█████████████████████████████████████████| 98.9M/98.9M [00:10<00:00, 10.3MB/s] Downloading: 100%|█████████████████████████████████████████████| 117/117 [00:00<00:00, 1.02MB/s] Downloading: 100%|█████████████████████████████████████████| 26.5M/26.5M [00:02<00:00, 10.0MB/s] Downloading: 100%|█████████████████████████████████████████████| 143/143 [00:00<00:00, 1.19MB/s] Downloading: 100%|██████████████████████████████████████████▉| 859M/859M [01:23<00:00, 10.7MB/s] Downloading: 100%|█████████████████████████████████████████████| 346/346 [00:00<00:00, 1.74MB/s] Downloading: 100%|█████████████████████████████████████████████| 309/309 [00:00<00:00, 2.64MB/s] Downloading: 100%|█████████████████████████████████████████| 1.36k/1.36k [00:00<00:00, 11.5MB/s] Downloading: 100%|█████████████████████████████████████████| 4.16k/4.16k [00:00<00:00, 12.3MB/s] Downloading: 100%|███████████████████████████████████████████| 329k/329k [00:00<00:00, 4.93MB/s] Downloading: 100%|█████████████████████████████████████████| 51.8M/51.8M [00:05<00:00, 9.35MB/s] Downloading: 100%|█████████████████████████████████████████████| 460/460 [00:00<00:00, 2.49MB/s] INFO:ChatTTS.core:Load from local: /host/chatTTS-ui/models/pzc163/chatTTS INFO:ChatTTS.core:use cuda:0 INFO:ChatTTS.core:vocos loaded. INFO:ChatTTS.core:dvae loaded. INFO:ChatTTS.core:gpt loaded. INFO:ChatTTS.core:decoder loaded. INFO:ChatTTS.core:tokenizer loaded. INFO:ChatTTS.core:All initialized. 启动:['0.0.0.0', '9966'] voice=2222,custom_voice=0 3%|█▍ | 10/384 [02:03<1:17:09, 12.38s/it] 4%|██▊ | 76/2048 [00:03<01:26, 22.92it/s] 推理时长: 127.39 秒 音频时长: 1.61 秒 voice=2222,custom_voice=0 18%|██████████████ | 70/384 [00:41<03:04, 1.70it/s] 32%|████████████████████████▏ | 661/2048 [00:38<01:21, 17.09it/s] 推理时长: 80.04 秒 音频时长: 14.09 秒 voice=4099,custom_voice=0 18%|██████████████ | 70/384 [00:03<00:17, 18.01it/s] 30%|██████████████████████▍ | 611/2048 [00:35<01:22, 17.43it/s] 推理时长: 39.09 秒 音频时长: 13.02 秒 voice=3333,custom_voice=0 18%|██████████████ | 70/384 [00:03<00:17, 17.98it/s] 30%|██████████████████████▏ | 606/2048 [00:34<01:22, 17.45it/s] 推理时长: 38.78 秒 音频时长: 12.92 秒 voice=7500,custom_voice=7500 WARNING:ChatTTS.core:Invalid characters found! : {'1', '4', '0'} 28%|█████████████████████▌ | 109/384 [00:44<01:52, 2.45it/s] 46%|██████████████████████████████████▌ | 945/2048 [01:03<01:13, 14.96it/s] 推理时长: 107.97 秒 音频时长: 20.15 秒 voice=2500,custom_voice=2500 WARNING:ChatTTS.core:Invalid characters found! : {'1', '4', '0'} 28%|█████████████████████▍ | 108/384 [00:06<00:15, 17.41it/s] 48%|███████████████████████████████████▊ | 978/2048 [01:06<01:12, 14.80it/s] 推理时长: 72.53 秒 音频时长: 20.85 秒 voice=2500,custom_voice=2500 WARNING:ChatTTS.core:Invalid characters found! : {':'} 8%|██████▏ | 31/384 [03:06<35:23, 6.02s/it] 15%|███████████ | 303/2048 [01:01<05:54, 4.93it/s] 推理时长: 248.17 秒 音频时长: 11.09 秒 ····

lin16303 commented 5 months ago

同样问题 6.txt ,1Torch was not compiled with flash attention,0.2版速度正常,0.3开始就极慢了

jianchang512 commented 5 months ago

源码部署下试试

xianglongwei commented 5 months ago

同样的问题,感觉GPU没有被调用。 image

jianchang512 commented 5 months ago

chat.load_models(source="local",local_path=CHATTTS_DIR) 里加个参数 compile=False

chat.load_models(source="local",local_path=CHATTTS_DIR,compile=False)

如果还是这样,就只能等 chatTTS官方优化了

cfso1234 commented 5 months ago

使用windows的源码部署方法,完美使用,3090生成的速度非常快,GPU也能100%工作了,compile=False参数一定要加,否则报错

lin16303 commented 5 months ago

打包版能否更新一下,谢谢

henrylaobai commented 5 months ago

0.6 版本 推理很慢,GPU 调用不到百分之10% {BD6B1BBA-9A87-44C3-AC38-E6B6904747F1}

jianchang512 commented 5 months ago

chat.load_models(source="local",local_path=CHATTTS_DIR) 里加个参数 compile=False

chat.load_models(source="local",local_path=CHATTTS_DIR,compile=False)

如果还是这样,就只能等 chatTTS官方优化了

chat.load_models(source="local",local_path=CHATTTS_DIR) 里加个参数 compile=False

chat.load_models(source="local",local_path=CHATTTS_DIR,compile=False)

如果还是这样,就只能等 chatTTS官方优化了

whldk commented 5 months ago

0.6 版本 推理很慢,GPU 调用不到百分之10% {BD6B1BBA-9A87-44C3-AC38-E6B6904747F1}

image

有没有可能是你自己没看到下面的

hotdogarea commented 5 months ago

使用windows的源码部署方法,完美使用,3090生成的速度非常快,GPU也能100%工作了,compile=False参数一定要加,否则报错

生成速度咋样,老哥

lin16303 commented 5 months ago

使用windows的源码部署方法,完美使用,3090生成的速度非常快,GPU也能100%工作了,compile=False参数一定要加,否则报错

生成速度咋样,老哥

4090 80its左右,大概1秒生成1秒音频,比GPT-SoVITS慢一半,edgetts接近200its