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.
#### What is your question? 进行快速训练的时候出现了报错 ModuleNotFoundError Traceback (most recent call last) Cell In[21], line 10 7 from modelscope.trainers import build_trainer 8 from modelscope.utils.constant import Tasks ---> 10 from funasr.datasets.ms_dataset...
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 执行: torchrun --nnodes 2 --node_rank 0 --nproc_per_node ${gpu_num} --master_addr ******* --master_port 1234...
## ❓ Questions and Help 如何通过FunASR的docker部署方法支持SenseVoice模型?
我运行alimeeting diarization 的 [run.sh](https://github.com/alibaba-damo-academy/FunASR/blob/main/egs/alimeeting/diarization/sond/run.sh)遇到错误,先是`utils/parse_options.sh`不存在。 我参考别的档案补上`../../../aishell/transformer/utils/`之后遇到run.sh: 34: utils/parse_options.sh: Syntax error: Bad for loop variable,不确定这个错误如何修復。 请问能协助我修正这个错误吗?
我这边做了一些尝试,初步定位到paraformer onnx-gpu耗时过长的原因: ## 1. predictor中的cif部分 可将其替换成 https://github.com/George0828Zhang/torch_cif (CIF的一种快速并行的实现方式,**但没确认是否等价于paraformer内部的实现**) ## 2. onnxruntime 中 CUDA Settings的问题: 1. cudnn_conv_algo_search 的默认配置是 EXHAUSTIVE,这个选项的默认配置是比较耗时的,尤其影响卷积操作(通过打印的日志可以发现耗时的部分全部集中在 decoder 部分的Conv_kernel_time) `"dur" :52419,"ts" :4481356,"ph" : "X","name" :"/decoder/decoders.X/self_attn/fsmn_block/Conv_kernel_time"`  2. 因此需要将配置修改为下面的 providers...
[www.funsound.cn](http://www.funsound.cn) 欢迎提出建议
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...
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 我用FunASR识别实时语音,由于那边推过来的流是通过WS推送PCM,每个包大小是234,然后用示例的funasr_wss_server.py去识别,vad和online效果不好。首先vad经常识别到的内容为[],导致fun_asr_online慢,然后fun_asr也执行的很慢,所以实时识别的数据推送出来的特别慢。 ### Before asking: 1. search the issues. 2. search...