bugfix

examples/aishell/branchformer/run.sh

@@ -109,6 +109,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
   echo "log_file: ${log_file}"
 
+  export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
   gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   torchrun \
   --nnodes 1 \

examples/aishell/conformer/run.sh

@@ -5,7 +5,7 @@ CUDA_VISIBLE_DEVICES="0,1"
 
 # general configuration
 feats_dir="../DATA" #feature output dictionary
-exp_dir="."
+exp_dir=`pwd`
 lang=zh
 token_type=char
 stage=0
@@ -109,6 +109,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
   echo "log_file: ${log_file}"
 
+  export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
   gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   torchrun \
   --nnodes 1 \
@@ -129,7 +130,7 @@ fi
 if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
   echo "stage 5: Inference"
 
-  if ${inference_device} == "cuda"; then
+  if [ ${inference_device} == "cuda" ]; then
       nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   else
       inference_batch_size=1
@@ -170,7 +171,7 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
           ++input="${_logdir}/keys.${JOB}.scp" \
           ++output_dir="${inference_dir}/${JOB}" \
           ++device="${inference_device}" \
-          ++batch_size="${inference_batch_size}"
+          ++batch_size="${inference_batch_size}" &> ${_logdir}/log.${JOB}.txt
         }&
 
     done

examples/aishell/e_branchformer/run.sh

@@ -109,6 +109,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
   echo "log_file: ${log_file}"
 
+  export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
   gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   torchrun \
   --nnodes 1 \

examples/aishell/paraformer/run.sh

@@ -109,6 +109,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
   echo "log_file: ${log_file}"
 
+  export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
   gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   torchrun \
   --nnodes 1 \

examples/aishell/transformer/run.sh

@@ -109,6 +109,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
   echo "log_file: ${log_file}"
 
+  export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
   gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   torchrun \
   --nnodes 1 \
@@ -129,7 +130,7 @@ fi
 if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
   echo "stage 5: Inference"
 
-  if ${inference_device} == "cuda"; then
+  if [ ${inference_device} == "cuda" ]; then
       nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   else
       inference_batch_size=1
@@ -141,7 +142,7 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
 
   for dset in ${test_sets}; do
 
-    inference_dir="${exp_dir}/exp/${model_dir}/${inference_checkpoint}/${dset}"
+    inference_dir="${exp_dir}/exp/${model_dir}/infer-${inference_checkpoint}/${dset}"
     _logdir="${inference_dir}/logdir"
 
     mkdir -p "${_logdir}"
@@ -154,7 +155,7 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
     done
     utils/split_scp.pl "${key_file}" ${split_scps}
 
-    gpuid_list_array=(${gpuid_list//,/ })
+    gpuid_list_array=(${CUDA_VISIBLE_DEVICES//,/ })
     for JOB in $(seq ${nj}); do
         {
           id=$((JOB-1))

examples/industrial_data_pretraining/bicif_paraformer/demo.py

@@ -11,8 +11,8 @@ model = AutoModel(model="damo/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k
                   vad_model_revision="v2.0.4",
                   punc_model="damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch",
                   punc_model_revision="v2.0.4",
-                  spk_model="damo/speech_campplus_sv_zh-cn_16k-common",
-                  spk_model_revision="v2.0.4",
+                  # spk_model="damo/speech_campplus_sv_zh-cn_16k-common",
+                  # spk_model_revision="v2.0.2",
                   )
 
 res = model.generate(input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_vad_punc_example.wav", batch_size_s=300, batch_size_threshold_s=60)

funasr/auto/auto_model.py

@@ -400,20 +400,20 @@ class AutoModel:
                     for res, vadsegment in zip(restored_data, vadsegments):
                         sentence_list.append({"start": vadsegment[0],\
                                                 "end": vadsegment[1],
-                                                "sentence": res['raw_text'],
+                                                "sentence": res['text'],
                                                 "timestamp": res['timestamp']})
                 elif self.spk_mode == 'punc_segment':
                     sentence_list = timestamp_sentence(punc_res[0]['punc_array'], \
                                                         result['timestamp'], \
-                                                        result['raw_text'])
+                                                        result['text'])
                 distribute_spk(sentence_list, sv_output)
                 result['sentence_info'] = sentence_list
             elif kwargs.get("sentence_timestamp", False):
                 sentence_list = timestamp_sentence(punc_res[0]['punc_array'], \
                                                         result['timestamp'], \
-                                                        result['raw_text'])
+                                                        result['text'])
                 result['sentence_info'] = sentence_list
-            del result['spk_embedding']
+            if "spk_embedding" in result: del result['spk_embedding']
                     
             result["key"] = key
             results_ret_list.append(result)

funasr/train_utils/trainer.py

@@ -279,7 +279,7 @@ class Trainer:
                     f"epoch: {epoch}/{self.max_epoch}, "
                     f"step: {batch_idx+1}/{len(self.dataloader_train)}, total: {self.batch_total}, "
                     f"(loss: {loss.detach().cpu().item():.3f}), "
-                    f"(lr: {lr}), "
+                    f"(lr: {lr:.3e}), "
                     f"{[(k, round(v.cpu().item(), 3)) for k, v in stats.items()]}, "
                     f"{speed_stats}, "
                     f"{gpu_info}"