MahmoudAshraf commited on
Commit
e311203
·
verified ·
1 Parent(s): 3bd2831

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -9
README.md CHANGED
@@ -176,7 +176,6 @@ pip install git+https://github.com/MahmoudAshraf97/ctc-forced-aligner.git
176
  ```
177
  ## Usage
178
 
179
- ```python
180
  from ctc_forced_aligner import (
181
  load_audio,
182
  load_alignment_model,
@@ -189,6 +188,7 @@ from ctc_forced_aligner import (
189
 
190
  audio_path = "your/audio/path"
191
  text_path = "your/text/path"
 
192
 
193
  audio_waveform = load_audio(audio_path, model.dtype, model.device)
194
  emissions, stride = generate_emissions(
@@ -202,10 +202,7 @@ text = "".join(line for line in lines).replace("\n", " ").strip()
202
  alignment_model, alignment_tokenizer, alignment_dictionary = load_alignment_model(
203
  device,
204
  dtype=torch.float16 if device == "cuda" else torch.float32,
205
- model_path="MahmoudAshraf/mms-300m-1130-forced-aligner"
206
  )
207
- # also compatible with other Wav2Vec2 Checkpoints such as
208
- # "jonatasgrosman/wav2vec2-large-xlsr-53-arabic"
209
 
210
 
211
  emissions, stride = generate_emissions(
@@ -213,14 +210,10 @@ emissions, stride = generate_emissions(
213
  )
214
 
215
 
216
- # romanization should be enabled when using multilingual models
217
- # it should be changed to `False` when using models that support the
218
- # native vocabulary of the text
219
-
220
  tokens_starred, text_starred = preprocess_text(
221
  text,
222
  romanize=True,
223
- language=langs_to_iso[language],
224
  )
225
 
226
 
 
176
  ```
177
  ## Usage
178
 
 
179
  from ctc_forced_aligner import (
180
  load_audio,
181
  load_alignment_model,
 
188
 
189
  audio_path = "your/audio/path"
190
  text_path = "your/text/path"
191
+ language = "iso" # ISO-639-3 Language code
192
 
193
  audio_waveform = load_audio(audio_path, model.dtype, model.device)
194
  emissions, stride = generate_emissions(
 
202
  alignment_model, alignment_tokenizer, alignment_dictionary = load_alignment_model(
203
  device,
204
  dtype=torch.float16 if device == "cuda" else torch.float32,
 
205
  )
 
 
206
 
207
 
208
  emissions, stride = generate_emissions(
 
210
  )
211
 
212
 
 
 
 
 
213
  tokens_starred, text_starred = preprocess_text(
214
  text,
215
  romanize=True,
216
+ language=language,
217
  )
218
 
219