PaddleSpeech
PaddleSpeech copied to clipboard
PaddlePaddle
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation a...
基于PaddleSpeech/examples/other/tts_finetune /tts3/的readme，在中英混合模型上，如果从BZNSYP中选出来3k条语音微调am模型，loss可以下降到0.7左右，且用微调模型合成语音，声音比较清晰，同样用aishell3的数据集的某个人的声音的多条数据微调，推理模型合成的声音也很清晰，没有沙沙的声音； 但是用上述方法，在thchs30上选了250个同一个人的语音进行微调，微调后推理模型合成的语音存在沙沙的声音，又从thchs30中选出1000条同一个人的音色的数据微调，微调后loss仍然在1.5左右，且推理合成的声音中存在沙沙的声音，但是能学到微调数据中的音色。 请问大佬们，上述是哪里出现了问题呢
根据`examples/wenetspeech/asr1/README.md`说明，对`asr1_chunk_conformer_u2pp_wenetspeech_ckpt_1.3.0.model.tar.gz`中的模型进行转化。 步骤如下： 1.进入转化脚本所在目录：`cd examples/wenetspeech/asr1/` 2.解压下载的模型文件压缩包至当前目录。 3.运行转化脚本：`./local/export.sh asr1_chunk_conformer_u2pp_wenetspeech_ckpt_1.3.0.model/model.yaml asr1_chunk_conformer_u2pp_wenetspeech_ckpt_1.3.0.model/exp/chunk_conformer_u2pp/checkpoints/avg_10 ./export.ji` 报错如下： ``` Traceback (most recent call last): File "/home/PaddleSpeech-r1.4/paddlespeech/s2t/exps/u2/bin//export.py", line 53, in main(config, args) File "/home/PaddleSpeech-r1.4/paddlespeech/s2t/exps/u2/bin//export.py", line 30, in main main_sp(config,...
采用fastspeech2_aishell3和pwgan_aishell3合成音频，指定spk_id后仍然出现多个人声音，而且有些字读不清晰，请问是什么原因呢？ 代码如下： source path.sh FLAGS_allocator_strategy=naive_best_fit \ FLAGS_fraction_of_gpu_memory_to_use=0.01 \ python3 ${BIN_DIR}/../synthesize_e2e.py \ --am fastspeech2_aishell3 \ --am_config fastspeech2_aishell3_ckpt_1.1.0/default.yaml \ --am_ckpt fastspeech2_aishell3_ckpt_1.1.0/snapshot_iter_96400.pdz \ --am_stat fastspeech2_aishell3_ckpt_1.1.0/speech_stats.npy \ --voc pwgan_aishell3 \ --voc_config pwg_aishell3_ckpt_0.5/default.yaml \ --voc_ckpt...
利用paddlespeech预训练模型进行声音分类模型调优 https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/docs/tutorial/cls/cls_tutorial.ipynb 1、发现同样的测试样本，特征提取也一样，但多次预测，每次预测的结果对应的分类类型都不一样 a、有尝试过固定随机种子，问题没解决 b、尝试导出静态模型进行推理，但predict时提示找不出lod_audio包（脚本参考https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/esc50/cls0/local/infer.sh） 问题情况：paddleaudio 1.01 load_audio包找不到，帮忙确认一下哪个版本有，谢谢 from paddleaudio.backends import soundfile_load as load_audio报错 ImportError Traceback (most recent call last) Cell In[12], line 21 19 import yaml 20 from paddle.audio.features...
audio_len should be 1D instead of 0D, which will raise list index out of range error in the following decode process ### PR types ### PR changes ### Describe
### PR types Others ### PR changes Others ### Describe Fixes a typo in README.md in demo
这是我的application.yaml protocol: 'http' engine_list: ['asr_python', 'text_python'] asr_python: model: 'conformer_wenetspeech' lang: 'zh' sample_rate: 16000 cfg_path: # [optional] ckpt_path: # [optional] decode_method: 'attention_rescoring' force_yes: True device: # set 'gpu:id' or 'cpu' text_python:...
RuntimeError: (NotFound) The kernel with key (GPU, Undefined(AnyLayout), int16) of kernel `pad3d` is not registered and fail to fallback to CPU one. Selected wrong DataType `int16`. Paddle support following DataTypes:...
Metadata
Owner
Metadata
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation a...