yiyun

期待数据集开源，希望团队中顶会

Building wheel for flash-attn (setup.py) ... - 一直卡在这里，好像很难装好

大佬我已经装上了，服务也启动起来了，但是推理不了，下面是具体信息 tts_data = { "text": response_text, # str.(required) text to be synthesized "text_lang": "zh", # str.(required) language of the text to be synthesized "ref_audio_path": "/home/GPT-SoVITS-main-V2-fast/output/guolimin/guolimin_0217_ref_anwei_0.wav", # str.(required) reference audio path...

nohup python api_v2.py -a 0.0.0.0 -p $TTS_1_PORTS -c GPT_SoVITS/configs/tts_infer_use1.yaml > nohup_logs/tts_1_v2.log 2>&1 & 使用的是api_v2.py，模型也是v2微调后的 下面是启动的脚本 custom: bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base device: cuda:0 is_half: true t2s_weights_path: GPT_weights_v2/chaonan_liu-e15.ckpt version: v2 vits_weights_path: SoVITS_weights_v2/chaonan_liu_e8_s48.pth...

好的感谢大佬，我试试

大佬我还是用这个脚本启动的，能用了，但是推理速度超级慢40秒一条，很奇怪 custom: bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base device: cuda:0 is_half: true t2s_weights_path: GPT_weights_v2/chaonan_liu-e15.ckpt version: v2 vits_weights_path: SoVITS_weights_v2/chaonan_liu_e8_s48.pth default: bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base device: cpu is_half: false t2s_weights_path: GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt version: v1 vits_weights_path:...