Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -11,7 +11,7 @@ from datetime import datetime
|
|
| 11 |
|
| 12 |
import assets
|
| 13 |
|
| 14 |
-
def sendToWhisper(audio_record, audio_upload,
|
| 15 |
results = []
|
| 16 |
|
| 17 |
audio = None
|
|
@@ -24,25 +24,14 @@ def sendToWhisper(audio_record, audio_upload, task, models_selected, language_to
|
|
| 24 |
|
| 25 |
audio = whisper.load_audio(audio)
|
| 26 |
audio = whisper.pad_or_trim(audio)
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
language = ""
|
| 36 |
-
prob = 0
|
| 37 |
-
if model_name in assets.lang_detect:
|
| 38 |
-
_, probs = model.detect_language(mel)
|
| 39 |
-
language = max(probs, key=probs.get)
|
| 40 |
-
prob = probs[language]
|
| 41 |
-
else:
|
| 42 |
-
language="en"
|
| 43 |
-
options = whisper.DecodingOptions(fp16 = False, without_timestamps=without_timestamps, task=task, language="en")
|
| 44 |
-
output_text = whisper.decode(model, mel, options)
|
| 45 |
-
results.append([model_name, output_text.text, language, str(prob), str((datetime.now() - start).total_seconds())])
|
| 46 |
return results
|
| 47 |
|
| 48 |
avail_models = whisper.available_models()
|
|
@@ -77,6 +66,7 @@ with gr.Blocks(css=assets.css) as demo:
|
|
| 77 |
gr.Markdown("## Output")
|
| 78 |
output = gr.Dataframe(headers=["Model", "Text", "Language", "Language Confidence","Time(s)"], label="Results", wrap=True)
|
| 79 |
|
| 80 |
-
submit.click(fn=sendToWhisper, inputs=[audio_record, audio_upload,
|
|
|
|
| 81 |
|
| 82 |
demo.launch()
|
|
|
|
| 11 |
|
| 12 |
import assets
|
| 13 |
|
| 14 |
+
def sendToWhisper(audio_record, audio_upload, without_timestamps):
|
| 15 |
results = []
|
| 16 |
|
| 17 |
audio = None
|
|
|
|
| 24 |
|
| 25 |
audio = whisper.load_audio(audio)
|
| 26 |
audio = whisper.pad_or_trim(audio)
|
| 27 |
+
|
| 28 |
+
start = datetime.now()
|
| 29 |
+
model = whisper.load_model(model_name)
|
| 30 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
| 31 |
+
# options = whisper.DecodingOptions(fp16 = False, without_timestamps=without_timestamps, task=task)
|
| 32 |
+
options = whisper.DecodingOptions()
|
| 33 |
+
output_text = whisper.decode(model, mel, options)
|
| 34 |
+
results.append([model_name, output_text.text, language, str(prob), str((datetime.now() - start).total_seconds())])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
return results
|
| 36 |
|
| 37 |
avail_models = whisper.available_models()
|
|
|
|
| 66 |
gr.Markdown("## Output")
|
| 67 |
output = gr.Dataframe(headers=["Model", "Text", "Language", "Language Confidence","Time(s)"], label="Results", wrap=True)
|
| 68 |
|
| 69 |
+
submit.click(fn=sendToWhisper, inputs=[audio_record, audio_upload, without_timestamps], outputs=output)
|
| 70 |
+
# [audio_record, audio_upload, task, models_selected, language_toggle, language_selected, without_timestamps], outputs=output)
|
| 71 |
|
| 72 |
demo.launch()
|