FunASR
FunASR copied to clipboard
A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.
In line 338 of funasr/models/emotion2vec/base.py we pass 4 arguments into compute_mask_indices (add_masks, seed, epoch, indices), which are not present in the function definition in fairseq_modules.py https://github.com/modelscope/FunASR/blob/1af68ba6ffc21d4dc3bd5f01cda656def97e361c/funasr/models/emotion2vec/base.py#L338
#### What is your question? For finetuning my model, should I prepare audio data less than 15s? I have lots of audios longer than 1 minute, should I split them...
求大神帮忙!!!! 问题描述: 我的服务器上有2台4090显卡,我开一个FunASR推理服务,RTF能到1000,我开启2个服务是,RTF能到2000。此时,CPU没满,显卡的显存和计算单元都没有满,然后IO,网卡。。。都没有满。但是我开到第三个服务时,RTF还只能到2000左右。 另外,我是走Nginx分发,配置没有问题,能上量。我的CPU是80C的。 使用的服务是:使用的官方的GPU镜像服务:https://github.com/modelscope/FunASR/blob/main/runtime/docs/benchmark_libtorch_cpp.md 使用的模型是:speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-torchscript 此外,我对服务和CPU做了绑核操作,让其服务质检不收干扰。使用推理的录音,都是30分钟左右的录音。 推理时,相关性能监控如下:  ![Uploading 2.png…]()
Notice: In order to resolve issues more efficiently, please raise issue following the template. (注意:为了更加高效率解决您遇到的问题,请按照模板提问,补充细节) ## 🐛 Bug ` def _load_audio_ffmpeg(file: str, sr: int = 16000): """ Open an audio...
想使用sencevoice模型进行asr,并且区分说话人。加载一直报错。 加载模型代码如下 from funasr import AutoModel from funasr.utils.postprocess_utils import rich_transcription_postprocess model = AutoModel(model="/data/asr/SenseVoice/iic/SenseVoiceSmall", vad_model="/data/asr/FunASR-main/model_zoo/fsmn-vad",punc_model="/data/asr/FunASR-main/model_zoo/ct-punc", vad_kwargs={"max_single_segment_time": 30000}, device="cuda:0", disable_update=True, spk_model="/data/asr/FunASR-main/model_zoo/cam" ) res = model.generate( input=f"/data/asr/FunASR-main/tests/voice.mp3", cache={}, language="auto", # "zn", "en", "yue",...
Notice: In order to resolve issues more efficiently, please raise issue following the template. (注意:为了更加高效率解决您遇到的问题,请按照模板提问,补充细节) ## ❓ Questions and Help ### Before asking: 1. search the issues. 2. search the...
## 🐛 Bug 尝试使用 finetune.sh 微调 seaco-paraformer ,微调后发现,每个epoch 的model.pt增大了很大,从原模型的800多M,增大到了2.44GB多,这是什么原因 ### To Reproduce bash finetune.sh Expected behavior 模型大小应保持一致 Environment OS (e.g., Linux): FunASR Version :1.1.5 PyTorch Version :2.3.1 How you installed...
Notice: In order to resolve issues more efficiently, please raise issue following the template. (注意:为了更加高效率解决您遇到的问题,请按照模板提问,补充细节) ## 🐛 Bug - `RuntimeError: PytorchStreamReader failed locating file data/2: file not found` - detail...