speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

funasr1.0 streaming

Browse Source
This commit is contained in:
游雁 2024-01-12 13:26:42 +08:00
parent 40d1f80030
commit bafd056c28
5 changed files with 28 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
examples/industrial_data_pretraining/fsmn_vad_streaming/demo.py
View File

@ -4,7 +4,8 @@
# MIT License (https://opensource.org/licenses/MIT) # MIT License (https://opensource.org/licenses/MIT)
from funasr import AutoModel from funasr import AutoModel
wav_file = "/Users/zhifu/funasr_github/test_local/asr_example.wav" wav_file = "https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/vad_example.wav"
chunk_size = 60000 # ms chunk_size = 60000 # ms
model = AutoModel(model="/Users/zhifu/Downloads/modelscope_models/speech_fsmn_vad_zh-cn-16k-common-streaming", model_revision="v2.0.0") model = AutoModel(model="/Users/zhifu/Downloads/modelscope_models/speech_fsmn_vad_zh-cn-16k-common-streaming", model_revision="v2.0.0")
@ -14,23 +15,23 @@ res = model(input=wav_file,
print(res) print(res)
#
# import soundfile import soundfile
# import os import os
#
# # wav_file = os.path.join(model.model_path, "example/vad_example.wav") wav_file = os.path.join(model.model_path, "example/vad_example.wav")
# speech, sample_rate = soundfile.read(wav_file) speech, sample_rate = soundfile.read(wav_file)
#
# chunk_stride = int(chunk_size * 16000 / 1000) chunk_stride = int(chunk_size * 16000 / 1000)
#
# cache = {} cache = {}
#
# for i in range(int(len((speech)-1)/chunk_stride+1)): for i in range(int(len((speech)-1)/chunk_stride+1)):
# speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride] speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride]
# is_final = i == int(len((speech)-1)/chunk_stride+1) is_final = i == int(len((speech)-1)/chunk_stride+1) - 1
# res = model(input=speech_chunk, res = model(input=speech_chunk,
# cache=cache, cache=cache,
# is_final=is_final, is_final=is_final,
# chunk_size=chunk_size, chunk_size=chunk_size,
# ) )
# print(res) print(res)

2
examples/industrial_data_pretraining/paraformer_streaming/demo.py
View File

@ -31,7 +31,7 @@ cache = {}
for i in range(int(len((speech)-1)/chunk_stride+1)): for i in range(int(len((speech)-1)/chunk_stride+1)):
speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride] speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride]
is_final = i == int(len((speech)-1)/chunk_stride+1) is_final = i == int(len((speech)-1)/chunk_stride+1) - 1
res = model(input=speech_chunk, res = model(input=speech_chunk,
cache=cache, cache=cache,
is_final=is_final, is_final=is_final,

4
funasr/frontends/wav_frontend.py
View File

@ -448,8 +448,8 @@ class WavFrontendOnline(nn.Module):
feats = torch.stack(cache["lfr_splice_cache"]) feats = torch.stack(cache["lfr_splice_cache"])
feats_lengths = torch.zeros(batch_size, dtype=torch.int) + feats.shape[1] feats_lengths = torch.zeros(batch_size, dtype=torch.int) + feats.shape[1]
feats, feats_lengths, _ = self.forward_lfr_cmvn(feats, feats_lengths, is_final, cache=cache) feats, feats_lengths, _ = self.forward_lfr_cmvn(feats, feats_lengths, is_final, cache=cache)
if is_final: # if is_final:
self.init_cache(cache) # self.init_cache(cache)
return feats, feats_lengths return feats, feats_lengths

4
funasr/models/fsmn_vad/model.py
View File

@ -12,6 +12,7 @@ from funasr.register import tables
from funasr.utils.load_utils import load_audio_text_image_video,extract_fbank from funasr.utils.load_utils import load_audio_text_image_video,extract_fbank
from funasr.utils.datadir_writer import DatadirWriter from funasr.utils.datadir_writer import DatadirWriter
from torch.nn.utils.rnn import pad_sequence from torch.nn.utils.rnn import pad_sequence
from funasr.train_utils.device_funcs import to_device
class VadStateMachine(Enum): class VadStateMachine(Enum):
kVadInStateStartPointNotDetected = 1 kVadInStateStartPointNotDetected = 1
@ -579,7 +580,8 @@ class FsmnVAD(nn.Module):
"cache": cache "cache": cache
} }
batch = to_device(batch, device=kwargs["device"])
segments_part, cache = self.forward(**batch) segments_part, cache = self.forward(**batch)
if segments_part: if segments_part:
for batch_num in range(0, batch_size): for batch_num in range(0, batch_size):

1
funasr/models/fsmn_vad_streaming/model.py
View File

@ -587,7 +587,6 @@ class FsmnVADStreaming(nn.Module):
"cache": cache["encoder"] "cache": cache["encoder"]
} }
segments_i = self.forward(**batch) segments_i = self.forward(**batch)
print(segments_i)
segments.extend(segments_i) segments.extend(segments_i)