FunASR
FunASR copied to clipboard
modelscope
A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.
# 0. 基于SenseVoiceSmall模型做微调训练时loss曲线出现负值 我在使用自己录的专业名词句子做微调训练的时候，看tensorboard中的train/loss曲线随着step的增加，这个loss曲线下降并趋于平稳，但是显示这个loss值是负数，感觉不对，请问你们有遇到过这样的问题吗？请问会是什么原因导致的呢？有大佬知道吗？ 我训练的数据量只有300多句话，训练的参数是大部分使用的fineturn.sh脚本中默认值 修改之后的参数如下： ```shell torchrun $DISTRIBUTED_ARGS \ ${train_tool} \ ++model="${model_name_or_model_dir}" \ ++train_data_set_list="${train_data}" \ ++valid_data_set_list="${val_data}" \ ++dataset_conf.data_split_num=1 \ ++dataset_conf.batch_sampler="BatchSampler" \ ++dataset_conf.batch_size=20000\ ++dataset_conf.sort_size=1024 \ ++dataset_conf.batch_type="token" \ ++dataset_conf.num_workers=16 \ ++train_conf.max_epoch=550...
## 🐛 Bug ### To Reproduce Steps to reproduce the behavior (**always include the command you ran**): 1. Run cmd '....' (fun) it@ai01:/opt/ai_users/abdurakhim/paraformer_train$ /home/it/anaconda3/envs/fun/bin/python /opt/ai_users/abdurakhim/paraformer_train/application/export_model.py funasr version: 1.2.6. Check update...
## 🐛 Bug 在使用pyinstaller打包程序后，运行报错 FileNotFoundError  ### To Reproduce ```python # main.py asr_model = AutoModel( model="paraformer-zh", vad_model="fsmn-vad", punc_model="ct-punc", disable_update=True, log_level="ERROR", disable_pbar=True, disable_log=True, ) asr_model.generate(input="音频路径") ``` ```bash pyinstaller -F main.py ```...
macbook pro m4，Python 3.11 ，funasr 1.2.6 from funasr import AutoModel model = AutoModel(model="Qwen/Qwen2-Audio-7B") audio_in = "/Users/hello/voice/output_000.mp3" prompt = "" res = model.generate(input=audio_in, prompt=prompt) print(res) 提示AssertionError: Qwen-Audio is not registered， 我通过这个代码查看funasr...
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...
项目地址：https://github.com/pika-online/Funsound2 官网地址：https://www.funsound.cn/  欢迎评测。
## ❓ Questions and Help #### What is your question? 运行最新版 onnxruntime cpu 版本 的2pass，发现在TpassUnit 等函数运行完后，内存仍有占用，且持续占用直到程序运行结束才会释放，这是为什么？我理解的话，不应该在Unit的时候就释放吗？
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 实时语音转写，开启逆文本正则化，十一点 会被转化为 10 一点，必现问题，11分，11秒等都能正常转换 ### Expected behavior 预期结果应该为11点 ### Environment - linux...
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 fun-onnx cuda-ep运行时快时慢
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...
Metadata
Owner
Metadata
A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.