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
游雁2e8dc0933f31
diff --git a/examples/aishell/branchformer/run.sh b/examples/aishell/branchformer/run.shindex 090226ba7..f4842d08c 100755--- a/examples/aishell/branchformer/run.sh+++ b/examples/aishell/branchformer/run.sh@@ -105,7 +105,8 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; thenecho "stage 4: ASR Training"mkdir -p ${exp_dir}/exp/${model_dir}- log_file="${exp_dir}/exp/${model_dir}/train.log.txt"+ current_time=$(date "+%Y-%m-%d_%H-%M")+ log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"echo "log_file: ${log_file}"gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')diff --git a/examples/aishell/e_branchformer/run.sh b/examples/aishell/e_branchformer/run.shindex 14cd64e5e..0855798c7 100755--- a/examples/aishell/e_branchformer/run.sh+++ b/examples/aishell/e_branchformer/run.sh@@ -105,7 +105,8 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; thenecho "stage 4: ASR Training"mkdir -p ${exp_dir}/exp/${model_dir}- log_file="${exp_dir}/exp/${model_dir}/train.log.txt"+ current_time=$(date "+%Y-%m-%d_%H-%M")+ log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"echo "log_file: ${log_file}"gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')diff --git a/examples/aishell/paraformer/run.sh b/examples/aishell/paraformer/run.shindex 90955f2b3..80f81b5f8 100755--- a/examples/aishell/paraformer/run.sh+++ b/examples/aishell/paraformer/run.sh@@ -105,7 +105,8 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; thenecho "stage 4: ASR Training"mkdir -p ${exp_dir}/exp/${model_dir}- log_file="${exp_dir}/exp/${model_dir}/train.log.txt"+ current_time=$(date "+%Y-%m-%d_%H-%M")+ log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"echo "log_file: ${log_file}"gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')diff --git a/examples/aishell/transformer/run.sh b/examples/aishell/transformer/run.shindex 90955f2b3..80f81b5f8 100755--- a/examples/aishell/transformer/run.sh+++ b/examples/aishell/transformer/run.sh@@ -105,7 +105,8 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; thenecho "stage 4: ASR Training"mkdir -p ${exp_dir}/exp/${model_dir}- log_file="${exp_dir}/exp/${model_dir}/train.log.txt"+ current_time=$(date "+%Y-%m-%d_%H-%M")+ log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"echo "log_file: ${log_file}"gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')diff --git a/funasr/train_utils/trainer.py b/funasr/train_utils/trainer.pyindex 10f7f80f0..cc7b215d2 100644--- a/funasr/train_utils/trainer.py+++ b/funasr/train_utils/trainer.py@@ -188,7 +188,7 @@ class Trainer:epoch (int): The current epoch number."""self.model.train()- pbar = tqdm(colour="blue", desc=f"Training Epoch: {epoch + 1}", total=len(self.dataloader_train),+ pbar = tqdm(colour="blue", desc=f"rank: {self.local_rank}, Training Epoch: {epoch + 1}", total=len(self.dataloader_train),dynamic_ncols=True)# Set the number of steps for gradient accumulation@@ -278,7 +278,7 @@ class Trainer:f"epoch: {epoch}/{self.max_epoch}, "f"step: {batch_idx}/{len(self.dataloader_train)}, total: {self.batch_total}, "f"(loss: {loss.detach().cpu().item():.3f}), "- f"{[(k, round(v.cpu().item(), 3)) for k, v in stats.items()]}"+ f"{[(k, round(v.cpu().item(), 3)) for k, v in stats.items()]}, "f"{speed_stats}, "f"{gpu_info}")@@ -307,7 +307,7 @@ class Trainer:"""self.model.eval()with torch.no_grad():- pbar = tqdm(colour="red", desc=f"Training Epoch: {epoch + 1}", total=len(self.dataloader_val),+ pbar = tqdm(colour="red", desc=f"rank: {self.local_rank}, Validation Epoch: {epoch + 1}", total=len(self.dataloader_val),dynamic_ncols=True)speed_stats = {}time5 = time.perf_counter()@@ -343,7 +343,7 @@ class Trainer:f"validation epoch: {epoch}/{self.max_epoch}, "f"step: {batch_idx}/{len(self.dataloader_val)}, "f"(loss: {loss.detach().cpu().item():.3f}), "- f"{[(k, round(v.cpu().item(), 3)) for k, v in stats.items()]}"+ f"{[(k, round(v.cpu().item(), 3)) for k, v in stats.items()]}, "f"{speed_stats}, ")pbar.set_description(description)