Merge branch 'alibaba-damo-academy:main' into add-offline-websocket-srv

egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online/infer.py

@@ -0,0 +1,39 @@
+import os
+import logging
+import torch
+import soundfile
+
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+from modelscope.utils.logger import get_logger
+
+logger = get_logger(log_level=logging.CRITICAL)
+logger.setLevel(logging.CRITICAL)
+
+os.environ["MODELSCOPE_CACHE"] = "./"
+inference_pipeline = pipeline(
+    task=Tasks.auto_speech_recognition,
+    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online',
+    model_revision='v1.0.4'
+)
+
+model_dir = os.path.join(os.environ["MODELSCOPE_CACHE"], "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online")
+speech, sample_rate = soundfile.read(os.path.join(model_dir, "example/asr_example.wav"))
+speech_length = speech.shape[0]
+
+sample_offset = 0
+chunk_size = [5, 10, 5] #[5, 10, 5] 600ms, [8, 8, 4] 480ms
+stride_size =  chunk_size[1] * 960
+param_dict = {"cache": dict(), "is_final": False, "chunk_size": chunk_size}
+final_result = ""
+
+for sample_offset in range(0, speech_length, min(stride_size, speech_length - sample_offset)):
+    if sample_offset + stride_size >= speech_length - 1:
+        stride_size = speech_length - sample_offset
+        param_dict["is_final"] = True
+    rec_result = inference_pipeline(audio_in=speech[sample_offset: sample_offset + stride_size],
+                                    param_dict=param_dict)
+    if len(rec_result) != 0:
+        final_result += rec_result['text'][0]
+        print(rec_result)
+print(final_result)

egs_modelscope/asr/paraformer/speech_paraformer_asr_nat-zh-cn-16k-common-vocab8404-online/infer.py

@@ -14,24 +14,26 @@ os.environ["MODELSCOPE_CACHE"] = "./"
 inference_pipeline = pipeline(
     task=Tasks.auto_speech_recognition,
     model='damo/speech_paraformer_asr_nat-zh-cn-16k-common-vocab8404-online',
-    model_revision='v1.0.2')
+    model_revision='v1.0.4'
+)
 
 model_dir = os.path.join(os.environ["MODELSCOPE_CACHE"], "damo/speech_paraformer_asr_nat-zh-cn-16k-common-vocab8404-online")
 speech, sample_rate = soundfile.read(os.path.join(model_dir, "example/asr_example.wav"))
 speech_length = speech.shape[0]
 
 sample_offset = 0
-step = 4800  #300ms
-param_dict = {"cache": dict(), "is_final": False}
+chunk_size = [8, 8, 4] #[5, 10, 5] 600ms, [8, 8, 4] 480ms
+stride_size =  chunk_size[1] * 960
+param_dict = {"cache": dict(), "is_final": False, "chunk_size": chunk_size}
 final_result = ""
 
-for sample_offset in range(0, speech_length, min(step, speech_length - sample_offset)):
-    if sample_offset + step >= speech_length - 1:
-        step = speech_length - sample_offset
+for sample_offset in range(0, speech_length, min(stride_size, speech_length - sample_offset)):
+    if sample_offset + stride_size >= speech_length - 1:
+        stride_size = speech_length - sample_offset
         param_dict["is_final"] = True
-    rec_result = inference_pipeline(audio_in=speech[sample_offset: sample_offset + step],
+    rec_result = inference_pipeline(audio_in=speech[sample_offset: sample_offset + stride_size],
                                     param_dict=param_dict)
-    if len(rec_result) != 0 and rec_result['text'] != "sil" and rec_result['text'] != "waiting_for_more_voice":
-        final_result += rec_result['text']
-    print(rec_result)
+    if len(rec_result) != 0:
+        final_result += rec_result['text'][0]
+        print(rec_result)
 print(final_result)

funasr/bin/asr_inference_paraformer_streaming.py

@@ -205,9 +205,12 @@ class Speech2Text:
         results = []
         cache_en = cache["encoder"]
         if speech.shape[1] < 16 * 60 and cache_en["is_final"]:
+            if cache_en["start_idx"] == 0:
+                return []
             cache_en["tail_chunk"] = True
             feats = cache_en["feats"]
             feats_len = torch.tensor([feats.shape[1]])
+            self.asr_model.frontend = None
             results = self.infer(feats, feats_len, cache)
             return results
         else:

funasr/models/encoder/sanm_encoder.py

@@ -380,7 +380,7 @@ class SANMEncoder(AbsEncoder):
         else:
             xs_pad = self.embed(xs_pad, cache)
         if cache["tail_chunk"]:
-            xs_pad = cache["feats"]
+            xs_pad = to_device(cache["feats"], device=xs_pad.device)
         else:
             xs_pad = self._add_overlap_chunk(xs_pad, cache)
         encoder_outs = self.encoders0(xs_pad, None, None, None, None)