csukuangfj commited on
Commit
f1df253
1 Parent(s): 7459972

minor fixes

Browse files
Files changed (2) hide show
  1. app.py +12 -1
  2. model.py +8 -9
app.py CHANGED
@@ -159,8 +159,14 @@ def process(
159
  info = f"""
160
  Wave duration : {duration: .3f} s <br/>
161
  Processing time: {end - start: .3f} s <br/>
162
- RTF: {end - start: .3f}/{duration: .3f} = {(end - start)/duration:.3f} <br/>
163
  """
 
 
 
 
 
 
164
  logging.info(info)
165
  logging.info(f"hyp:\n{hyp}")
166
 
@@ -171,12 +177,17 @@ title = "# Automatic Speech Recognition with Next-gen Kaldi"
171
  description = """
172
  This space shows how to do automatic speech recognition with Next-gen Kaldi.
173
 
 
 
174
  See more information by visiting the following links:
175
 
176
  - <https://github.com/k2-fsa/icefall>
177
  - <https://github.com/k2-fsa/sherpa>
178
  - <https://github.com/k2-fsa/k2>
179
  - <https://github.com/lhotse-speech/lhotse>
 
 
 
180
  """
181
 
182
  # css style is copied from
 
159
  info = f"""
160
  Wave duration : {duration: .3f} s <br/>
161
  Processing time: {end - start: .3f} s <br/>
162
+ RTF: {end - start: .3f}/{duration: .3f} = {rtf:.3f} <br/>
163
  """
164
+ if rtf > 1:
165
+ info += (
166
+ f"<br/>We are loading the model for the first run. "
167
+ "Please run again to measure the real RTF.<br/>"
168
+ )
169
+
170
  logging.info(info)
171
  logging.info(f"hyp:\n{hyp}")
172
 
 
177
  description = """
178
  This space shows how to do automatic speech recognition with Next-gen Kaldi.
179
 
180
+ It is running on CPU within a docker container provided by Hugging Face.
181
+
182
  See more information by visiting the following links:
183
 
184
  - <https://github.com/k2-fsa/icefall>
185
  - <https://github.com/k2-fsa/sherpa>
186
  - <https://github.com/k2-fsa/k2>
187
  - <https://github.com/lhotse-speech/lhotse>
188
+
189
+ If you want to deploy it locally, please see
190
+ <https://k2-fsa.github.io/sherpa/>
191
  """
192
 
193
  # css style is copied from
model.py CHANGED
@@ -81,7 +81,7 @@ def _get_aishell2_pretrained_model(repo_id: str) -> OfflineAsr:
81
  "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12", # noqa
82
  # context-size 2
83
  "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12", # noqa
84
- ]
85
 
86
  nn_model_filename = _get_nn_model_filename(
87
  repo_id=repo_id,
@@ -102,12 +102,11 @@ def _get_aishell2_pretrained_model(repo_id: str) -> OfflineAsr:
102
  def _get_gigaspeech_pre_trained_model(repo_id: str) -> OfflineAsr:
103
  assert repo_id in [
104
  "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
105
- ]
106
 
107
  nn_model_filename = _get_nn_model_filename(
108
- # It is converted from https://huggingface.co/wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2 # noqa
109
- repo_id="csukuangfj/icefall-asr-gigaspeech-pruned-transducer-stateless2", # noqa
110
- filename="cpu_jit-epoch-29-avg-11-torch-1.10.0.pt",
111
  )
112
  bpe_model_filename = _get_bpe_model_filename(repo_id=repo_id)
113
 
@@ -124,7 +123,7 @@ def _get_gigaspeech_pre_trained_model(repo_id: str) -> OfflineAsr:
124
  def _get_librispeech_pre_trained_model(repo_id: str) -> OfflineAsr:
125
  assert repo_id in [
126
  "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13", # noqa
127
- ]
128
 
129
  nn_model_filename = _get_nn_model_filename(
130
  repo_id=repo_id,
@@ -145,7 +144,7 @@ def _get_librispeech_pre_trained_model(repo_id: str) -> OfflineAsr:
145
  def _get_wenetspeech_pre_trained_model(repo_id: str):
146
  assert repo_id in [
147
  "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
148
- ]
149
 
150
  nn_model_filename = _get_nn_model_filename(
151
  repo_id=repo_id,
@@ -166,7 +165,7 @@ def _get_wenetspeech_pre_trained_model(repo_id: str):
166
  def _get_tal_csasr_pre_trained_model(repo_id: str):
167
  assert repo_id in [
168
  "luomingshuang/icefall_asr_tal-csasr_pruned_transducer_stateless5",
169
- ]
170
 
171
  nn_model_filename = _get_nn_model_filename(
172
  repo_id=repo_id,
@@ -187,7 +186,7 @@ def _get_tal_csasr_pre_trained_model(repo_id: str):
187
  def _get_alimeeting_pre_trained_model(repo_id: str):
188
  assert repo_id in [
189
  "luomingshuang/icefall_asr_alimeeting_pruned_transducer_stateless2",
190
- ]
191
 
192
  nn_model_filename = _get_nn_model_filename(
193
  repo_id=repo_id,
 
81
  "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12", # noqa
82
  # context-size 2
83
  "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12", # noqa
84
+ ], repo_id
85
 
86
  nn_model_filename = _get_nn_model_filename(
87
  repo_id=repo_id,
 
102
  def _get_gigaspeech_pre_trained_model(repo_id: str) -> OfflineAsr:
103
  assert repo_id in [
104
  "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
105
+ ], repo_id
106
 
107
  nn_model_filename = _get_nn_model_filename(
108
+ repo_id=repo_id,
109
+ filename="cpu_jit-iter-3488000-avg-20.pt",
 
110
  )
111
  bpe_model_filename = _get_bpe_model_filename(repo_id=repo_id)
112
 
 
123
  def _get_librispeech_pre_trained_model(repo_id: str) -> OfflineAsr:
124
  assert repo_id in [
125
  "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13", # noqa
126
+ ], repo_id
127
 
128
  nn_model_filename = _get_nn_model_filename(
129
  repo_id=repo_id,
 
144
  def _get_wenetspeech_pre_trained_model(repo_id: str):
145
  assert repo_id in [
146
  "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
147
+ ], repo_id
148
 
149
  nn_model_filename = _get_nn_model_filename(
150
  repo_id=repo_id,
 
165
  def _get_tal_csasr_pre_trained_model(repo_id: str):
166
  assert repo_id in [
167
  "luomingshuang/icefall_asr_tal-csasr_pruned_transducer_stateless5",
168
+ ], repo_id
169
 
170
  nn_model_filename = _get_nn_model_filename(
171
  repo_id=repo_id,
 
186
  def _get_alimeeting_pre_trained_model(repo_id: str):
187
  assert repo_id in [
188
  "luomingshuang/icefall_asr_alimeeting_pruned_transducer_stateless2",
189
+ ], repo_id
190
 
191
  nn_model_filename = _get_nn_model_filename(
192
  repo_id=repo_id,