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
train finetune demo
游雁6aba0e3b5689
diff --git a/.gitignore b/.gitignoreindex 6bdfd5d06..adf2937f6 100644--- a/.gitignore+++ b/.gitignore@@ -24,3 +24,4 @@ samplesoutputs*emotion2vec*GPT-SoVITS*+modelscope_modelsdiff --git a/examples/aishell/paraformer/demo_train_or_finetune.sh b/examples/aishell/paraformer/demo_train_or_finetune.shindex 4c95dc763..06607c707 100644--- a/examples/aishell/paraformer/demo_train_or_finetune.sh+++ b/examples/aishell/paraformer/demo_train_or_finetune.sh@@ -17,7 +17,7 @@ data_dir="/Users/zhifu/funasr1.0/data/list"train_data="${data_dir}/train.jsonl"val_data="${data_dir}/val.jsonl"-tokens="${data_dir}/tokens.jsonl"+tokens="${data_dir}/tokens.json"cmvn_file="${data_dir}/am.mvn"# exp output dir@@ -45,7 +45,7 @@ torchrun \++dataset_conf.batch_size=32 \++dataset_conf.batch_type="example" \++dataset_conf.num_workers=4 \-++train_conf.max_epoch=20 \+++train_conf.max_epoch=150 \++optim_conf.lr=0.0002 \++init_param="${init_param}" \++output_dir="${output_dir}" &> ${log_file}diff --git a/examples/industrial_data_pretraining/paraformer/demo.sh b/examples/industrial_data_pretraining/paraformer/demo.shdeleted file mode 100644index 477aee6d2..000000000--- a/examples/industrial_data_pretraining/paraformer/demo.sh+++ /dev/null@@ -1,29 +0,0 @@--# method1, inference from model hub--model="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"-model_revision="v2.0.4"--python -m funasr.bin.inference \-+model=${model} \-+model_revision=${model_revision} \-+input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav" \-+output_dir="./outputs/debug" \-+device="cpu" \---# method2, inference from local model--#python -m funasr.bin.inference \-#--config-path="/Users/zhifu/funasr_github/test_local/funasr_cli_egs" \-#--config-name="config.yaml" \-#++init_param="/Users/zhifu/funasr_github/test_local/funasr_cli_egs/model.pt" \-#++tokenizer_conf.token_list="/Users/zhifu/funasr_github/test_local/funasr_cli_egs/tokens.txt" \-#++frontend_conf.cmvn_file="/Users/zhifu/funasr_github/test_local/funasr_cli_egs/am.mvn" \-#++input="data/wav.scp" \-#++output_dir="./outputs/debug" \-#++device="cuda" \----diff --git a/examples/industrial_data_pretraining/paraformer/finetune_from_model_hub.sh b/examples/industrial_data_pretraining/paraformer/finetune.shsimilarity index 86%rename from examples/industrial_data_pretraining/paraformer/finetune_from_model_hub.shrename to examples/industrial_data_pretraining/paraformer/finetune.shindex 0ccb83a12..21b29b6b1 100644--- a/examples/industrial_data_pretraining/paraformer/finetune_from_model_hub.sh+++ b/examples/industrial_data_pretraining/paraformer/finetune.sh@@ -1,6 +1,8 @@# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.# MIT License (https://opensource.org/licenses/MIT)+# method1, finetune from model hub+# which gpu to train or finetuneexport CUDA_VISIBLE_DEVICES="0,1"gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')@@ -30,8 +32,8 @@ torchrun \--nnodes 1 \--nproc_per_node ${gpu_num} \funasr/bin/train.py \-+model="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch" \-+model_revision="v2.0.4" \+++model="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch" \+++model_revision="v2.0.4" \++train_data_set_list="${train_data}" \++valid_data_set_list="${val_data}" \++dataset_conf.batch_size=32 \diff --git a/examples/industrial_data_pretraining/paraformer/finetune_from_local_model.sh b/examples/industrial_data_pretraining/paraformer/finetune_from_local.shsimilarity index 91%rename from examples/industrial_data_pretraining/paraformer/finetune_from_local_model.shrename to examples/industrial_data_pretraining/paraformer/finetune_from_local.shindex 2e0b8eeea..5b4d99106 100644--- a/examples/industrial_data_pretraining/paraformer/finetune_from_local_model.sh+++ b/examples/industrial_data_pretraining/paraformer/finetune_from_local.sh@@ -1,8 +1,12 @@# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.# MIT License (https://opensource.org/licenses/MIT)+# method2, finetune from local model++workspace=`pwd`+# download model-local_path_root=../modelscope_models+local_path_root=${workspace}/modelscope_modelsmkdir -p ${local_path_root}local_path=${local_path_root}/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorchgit clone https://www.modelscope.cn/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch.git ${local_path}@@ -24,15 +28,14 @@ data_dir="/Users/zhifu/funasr1.0/data/list"train_data="${data_dir}/train.jsonl"val_data="${data_dir}/val.jsonl"-tokens="${local_path}/tokens.jsonl"+tokens="${local_path}/tokens.json"cmvn_file="${local_path}/am.mvn"# exp output diroutput_dir="/Users/zhifu/exp"log_file="${output_dir}/log.txt"-workspace=`pwd`-config="${local_path}/config.yaml"+config="config.yaml"init_param="${local_path}/model.pt"@@ -43,7 +46,7 @@ torchrun \--nnodes 1 \--nproc_per_node ${gpu_num} \../../../funasr/bin/train.py \---config-path "${workspace}/conf" \+--config-path "${local_path}" \--config-name "${config}" \++train_data_set_list="${train_data}" \++valid_data_set_list="${val_data}" \diff --git a/examples/industrial_data_pretraining/paraformer/infer.sh b/examples/industrial_data_pretraining/paraformer/infer.shnew file mode 100644index 000000000..a1975eac8--- /dev/null+++ b/examples/industrial_data_pretraining/paraformer/infer.sh@@ -0,0 +1,21 @@+# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.+# MIT License (https://opensource.org/licenses/MIT)++# method1, inference from model hub++# for more input type, please ref to readme.md+input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav"++output_dir="./outputs/debug"++model="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"+model_revision="v2.0.4"++device="cuda:0" # "cuda:0" for gpu0, "cuda:1" for gpu1, "cpu"++python -m funasr.bin.inference \+++model=${model} \+++model_revision=${model_revision} \+++input="${input}" \+++output_dir="${output_dir}" \+++device="${device}" \diff --git a/examples/industrial_data_pretraining/paraformer/infer_from_local.sh b/examples/industrial_data_pretraining/paraformer/infer_from_local.shnew file mode 100644index 000000000..06c9ca620--- /dev/null+++ b/examples/industrial_data_pretraining/paraformer/infer_from_local.sh@@ -0,0 +1,39 @@+# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.+# MIT License (https://opensource.org/licenses/MIT)++# method2, inference from local model++# for more input type, please ref to readme.md+input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav"++output_dir="./outputs/debug"++workspace=`pwd`++# download model+local_path_root=${workspace}/modelscope_models+mkdir -p ${local_path_root}+local_path=${local_path_root}/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch+git clone https://www.modelscope.cn/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch.git ${local_path}++device="cuda:0" # "cuda:0" for gpu0, "cuda:1" for gpu1, "cpu"++tokens="${local_path}/tokens.json"+cmvn_file="${local_path}/am.mvn"++config="config.yaml"+init_param="${local_path}/model.pt"++python -m funasr.bin.inference \+--config-path "${local_path}" \+--config-name "${config}" \+++init_param="${init_param}" \+++tokenizer_conf.token_list="${tokens}" \+++frontend_conf.cmvn_file="${cmvn_file}" \+++input="${input}" \+++output_dir="${output_dir}" \+++device="${device}" \++++