Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
funasr tables
游雁98e2c546a089
diff --git a/docs/tutorial/README_zh.md b/docs/tutorial/README_zh.mdindex afef070bf..324227531 100644--- a/docs/tutorial/README_zh.md+++ b/docs/tutorial/README_zh.md@@ -442,22 +442,21 @@ print(result)### 查看注册表-```python+```plaintextfrom funasr.register import tablestables.print()```-支持查看指定类型的注册表:`tables.print("model")`-+支持查看指定类型的注册表:\`tables.print("model")\`-### 注册新模型+### 注册模型```pythonfrom funasr.register import tables[email protected]("model_classes", "MinMo_S2T")-class MinMo_S2T(nn.Module):[email protected]("model_classes", "SenseVoiceSmall")+class SenseVoiceSmall(nn.Module):def __init__(*args, **kwargs):...@@ -479,10 +478,14 @@ class MinMo_S2T(nn.Module):```-然后在config.yaml中指定新注册模型+在需要注册的类名前加上 `@tables.register("model_classes","SenseVoiceSmall")`,即可完成注册,类需要实现有:__init__,forward,inference方法。++完整代码:[https://github.com/modelscope/FunASR/blob/main/funasr/models/sense\_voice/model.py#L443](https://github.com/modelscope/FunASR/blob/main/funasr/models/sense_voice/model.py#L443)-```yaml-model: MinMo_S2T+注册完成后,在config.yaml中指定新注册模型,即可实现对模型的定义++```python+model: SenseVoiceSmallmodel_conf:...```diff --git a/docs/tutorial/Tables_zh.md b/docs/tutorial/Tables_zh.mdindex ec64baf1e..967c625be 100644--- a/docs/tutorial/Tables_zh.md+++ b/docs/tutorial/Tables_zh.md@@ -1,4 +1,4 @@-# FunASR-1.x.x 注册模型教程+# FunASR-1.x.x 注册模型教程1.0版本的设计初衷是【**让模型集成更简单**】,核心feature为注册表与AutoModel:@@ -11,7 +11,7 @@* 统一学术与工业模型推理训练脚本;-+# 快速上手@@ -89,14 +89,14 @@ model = AutoModel(model=[str], device=[str], ncpu=[int], output_dir=[str], batchres = model.generate(input=[str], output_dir=[str])```-* wav文件路径, 例如: asr\_example.wav--* pcm文件路径, 例如: asr\_example.pcm,此时需要指定音频采样率fs(默认为16000)--* 音频字节数流,例如:麦克风的字节数数据--* wav.scp,kaldi 样式的 wav 列表 (`wav_id \t wav_path`), 例如:-+* * wav文件路径, 例如: asr\_example.wav++ * pcm文件路径, 例如: asr\_example.pcm,此时需要指定音频采样率fs(默认为16000)++ * 音频字节数流,例如:麦克风的字节数数据++ * wav.scp,kaldi 样式的 wav 列表 (`wav_id \t wav_path`), 例如:+```plaintextasr_example1 ./audios/asr_example1.wav@@ -121,76 +121,82 @@ asr_example2 ./audios/asr_example2.wav## 模型资源目录-+++**模型链接为:**[https://www.modelscope.cn/models/iic/SenseVoiceSmall/files](https://www.modelscope.cn/models/iic/SenseVoiceSmall/files)**配置文件**:config.yaml```yaml-model: SenseVoiceLarge-model_conf:- lsm_weight: 0.1- length_normalized_loss: true- activation_checkpoint: true- sos: <|startoftranscript|>- eos: <|endoftext|>- downsample_rate: 4- use_padmask: true--encoder: SenseVoiceEncoder+encoder: SenseVoiceEncoderSmallencoder_conf:- input_size: 128- attention_heads: 20- linear_units: 1280- num_blocks: 32- dropout_rate: 0.1- positional_dropout_rate: 0.1- attention_dropout_rate: 0.1- kernel_size: 31- sanm_shfit: 0- att_type: self_att_fsmn_sdpa- downsample_rate: 4- use_padmask: true- max_position_embeddings: 2048- rope_theta: 10000--frontend: WhisperFrontend-frontend_conf:- fs: 16000- n_mels: 128- do_pad_trim: false- filters_path: null+ output_size: 512+ attention_heads: 4+ linear_units: 2048+ num_blocks: 50+ tp_blocks: 20+ dropout_rate: 0.1+ positional_dropout_rate: 0.1+ attention_dropout_rate: 0.1+ input_layer: pe+ pos_enc_class: SinusoidalPositionEncoder+ normalize_before: true+ kernel_size: 11+ sanm_shfit: 0+ selfattention_layer_type: sanm+++model: SenseVoiceSmall+model_conf:+ length_normalized_loss: true+ sos: 1+ eos: 2+ ignore_id: -1-tokenizer: SenseVoiceTokenizer+tokenizer: SentencepiecesTokenizertokenizer_conf:- vocab_path: null- is_multilingual: true- num_languages: 8749+ bpemodel: null+ unk_symbol: <unk>+ split_with_space: true++frontend: WavFrontend+frontend_conf:+ fs: 16000+ window: hamming+ n_mels: 80+ frame_length: 25+ frame_shift: 10+ lfr_m: 7+ lfr_n: 6+ cmvn_file: null-dataset: SenseVoiceDataset++dataset: SenseVoiceCTCDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: BatchSampler+ batch_sampler: EspnetStyleBatchSampler+ data_split_num: 32batch_type: token- batch_size: 12000- sort_size: 64+ batch_size: 14000max_token_length: 2000min_token_length: 60max_source_length: 2000min_source_length: 60- max_target_length: 150+ max_target_length: 200min_target_length: 0shuffle: truenum_workers: 4sos: ${model_conf.sos}eos: ${model_conf.eos}IndexDSJsonl: IndexDSJsonl+ retry: 20train_conf:accum_grad: 1grad_clip: 5- max_epoch: 5- keep_nbest_models: 200- avg_nbest_model: 200+ max_epoch: 20+ keep_nbest_models: 10+ avg_nbest_model: 10log_interval: 100resume: truevalidate_interval: 10000@@ -198,11 +204,10 @@ train_conf:optim: adamwoptim_conf:- lr: 2.5e-05-+ lr: 0.00002scheduler: warmuplrscheduler_conf:- warmup_steps: 20000+ warmup_steps: 25000```@@ -222,8 +227,8 @@ scheduler_conf:"file_path_metas": {"init_param":"model.pt","config":"config.yaml",- "tokenizer_conf": {"vocab_path": "tokens.tiktoken"},- "frontend_conf":{"filters_path": "mel_filters.npz"}}+ "tokenizer_conf": {"bpemodel": "chn_jpn_yue_eng_ko_spectok.bpe.model"},+ "frontend_conf":{"cmvn_file": "am.mvn"}}}```@@ -231,22 +236,21 @@ scheduler_conf:### 查看注册表-```python+```plaintextfrom funasr.register import tablestables.print()```-支持查看指定类型的注册表:`tables.print("model")`-+支持查看指定类型的注册表:\`tables.print("model")\`-### 新注册+### 注册模型```pythonfrom funasr.register import tables[email protected]("model_classes", "MinMo_S2T")-class MinMo_S2T(nn.Module):[email protected]("model_classes", "SenseVoiceSmall")+class SenseVoiceSmall(nn.Module):def __init__(*args, **kwargs):...@@ -268,10 +272,14 @@ class MinMo_S2T(nn.Module):```-在config.yaml中指定新注册模型+在需要注册的类名前加上 `@tables.register("model_classes","SenseVoiceSmall")`,即可完成注册,类需要实现有:__init__,forward,inference方法。-```yaml-model: MinMo_S2T+完整代码:[https://github.com/modelscope/FunASR/blob/main/funasr/models/sense\_voice/model.py#L443](https://github.com/modelscope/FunASR/blob/main/funasr/models/sense_voice/model.py#L443)++注册完成后,在config.yaml中指定新注册模型,即可实现对模型的定义++```python+model: SenseVoiceSmallmodel_conf:...```