Merge pull request #939 from alibaba-damo-academy/dev_sxfix

Bug fix

funasr/bin/asr_inference_launch.py

@@ -415,7 +415,7 @@ def inference_paraformer(
                         ibest_writer["rtf"][key] = rtf_cur
 
                     if text is not None:
-                        if use_timestamp and timestamp is not None:
+                        if use_timestamp and timestamp is not None and len(timestamp):
                             postprocessed_result = postprocess_utils.sentence_postprocess(token, timestamp)
                         else:
                             postprocessed_result = postprocess_utils.sentence_postprocess(token)
@@ -427,7 +427,7 @@ def inference_paraformer(
                         else:
                             text_postprocessed, word_lists = postprocessed_result[0], postprocessed_result[1]
                         item = {'key': key, 'value': text_postprocessed}
-                        if timestamp_postprocessed != "":
+                        if timestamp_postprocessed != "" or len(timestamp) == 0:
                             item['timestamp'] = timestamp_postprocessed
                         asr_result_list.append(item)
                         finish_count += 1
@@ -692,7 +692,7 @@ def inference_paraformer_vad_punc(
             text, token, token_int = result[0], result[1], result[2]
             time_stamp = result[4] if len(result[4]) > 0 else None
 
-            if use_timestamp and time_stamp is not None:
+            if use_timestamp and time_stamp is not None and len(time_stamp):
                 postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
             else:
                 postprocessed_result = postprocess_utils.sentence_postprocess(token)
@@ -717,7 +717,7 @@ def inference_paraformer_vad_punc(
             item = {'key': key, 'value': text_postprocessed_punc}
             if text_postprocessed != "":
                 item['text_postprocessed'] = text_postprocessed
-            if time_stamp_postprocessed != "":
+            if time_stamp_postprocessed != "" or len(time_stamp) == 0:
                 item['time_stamp'] = time_stamp_postprocessed
 
             item['sentences'] = time_stamp_sentence(punc_id_list, time_stamp_postprocessed, text_postprocessed)

funasr/export/export_model.py

@@ -254,7 +254,7 @@ class ModelExport:
             if not os.path.exists(quant_model_path):
                 onnx_model = onnx.load(model_path)
                 nodes = [n.name for n in onnx_model.graph.node]
-                nodes_to_exclude = [m for m in nodes if 'output' in m]
+                nodes_to_exclude = [m for m in nodes if 'output' in m or 'bias_encoder' in m  or 'bias_decoder' in m]
                 quantize_dynamic(
                     model_input=model_path,
                     model_output=quant_model_path,

funasr/runtime/python/onnxruntime/demo_contextual_paraformer.py

@@ -5,7 +5,7 @@ model_dir = "./export/damo/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-
 model = ContextualParaformer(model_dir, batch_size=1)
 
 wav_path = ['{}/.cache/modelscope/hub/damo/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404/example/asr_example.wav'.format(Path.home())]
-hotwords = '随机热词 各种热词 魔搭 阿里巴巴'
+hotwords = '随机热词 各种热词 魔搭 阿里巴巴 仏'
 
 result = model(wav_path, hotwords)
 print(result)

funasr/runtime/python/onnxruntime/funasr_onnx/paraformer_bin.py

@@ -314,7 +314,14 @@ class ContextualParaformer(Paraformer):
         hotwords_length = torch.Tensor(hotwords_length).to(torch.int32)
         # hotwords.append('<s>')
         def word_map(word):
-            return torch.tensor([self.vocab[i] for i in word])
+            hotwords = []
+            for c in word:
+                if c not in self.vocab.keys():
+                    hotwords.append(8403)
+                    logging.warning("oov character {} found in hotword {}, replaced by <unk>".format(c, word))
+                else:
+                    hotwords.append(self.vocab[c])
+            return torch.tensor(hotwords)
         hotword_int = [word_map(i) for i in hotwords]
         # import pdb; pdb.set_trace()
         hotword_int.append(torch.tensor([1]))