From 94cb66dbb9ae12e044a41fb8a3d84e1835ee7e7b Mon Sep 17 00:00:00 2001
From: zhifu gao <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 02 三月 2023 20:20:10 +0800
Subject: [PATCH] Merge pull request #177 from alibaba-damo-academy/dev_timestamp
---
funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py | 13 +++++++++----
1 files changed, 9 insertions(+), 4 deletions(-)
diff --git a/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py b/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
index 8af3474..9b8a67b 100644
--- a/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
+++ b/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
@@ -14,7 +14,7 @@
read_yaml)
from .utils.postprocess_utils import sentence_postprocess
from .utils.frontend import WavFrontend
-from funasr.utils.timestamp_tools import time_stamp_lfr6_pl
+from .utils.timestamp_utils import time_stamp_lfr6_onnx
logging = get_logger()
@@ -41,17 +41,16 @@
)
self.ort_infer = OrtInferSession(model_file, device_id)
self.batch_size = batch_size
+ self.plot = True
def __call__(self, wav_content: Union[str, np.ndarray, List[str]], **kwargs) -> List:
waveform_list = self.load_data(wav_content, self.frontend.opts.frame_opts.samp_freq)
waveform_nums = len(waveform_list)
-
asr_res = []
for beg_idx in range(0, waveform_nums, self.batch_size):
res = {}
end_idx = min(waveform_nums, beg_idx + self.batch_size)
feats, feats_len = self.extract_feat(waveform_list[beg_idx:end_idx])
-
try:
outputs = self.infer(feats, feats_len)
am_scores, valid_token_lens = outputs[0], outputs[1]
@@ -68,11 +67,17 @@
preds, raw_token = self.decode(am_scores, valid_token_lens)[0]
res['preds'] = preds
if us_cif_peak is not None:
- timestamp = time_stamp_lfr6_pl(us_alphas, us_cif_peak, copy.copy(raw_token), log=False)
+ timestamp, timestamp_total = time_stamp_lfr6_onnx(us_cif_peak, copy.copy(raw_token))
res['timestamp'] = timestamp
+ if self.plot:
+ self.plot_wave_timestamp(waveform_list[0], timestamp_total)
asr_res.append(res)
return asr_res
+ def plot_wave_timestamp(self, wav, text_timestamp):
+ # TODO: Plot the wav and timestamp results with matplotlib
+ import pdb; pdb.set_trace()
+
def load_data(self,
wav_content: Union[str, np.ndarray, List[str]], fs: int = None) -> List:
def load_wav(path: str) -> np.ndarray:
--
Gitblit v1.9.1