RamAnanth1 commited on
Commit
4094da1
·
1 Parent(s): 49faccd

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +19 -13
app.py CHANGED
@@ -19,9 +19,15 @@ whisper_model = pipeline(
19
  device=device,
20
  )
21
 
22
- # def transcribe(audio):
23
- # text = whisper_model(audio)["text"]
24
- # return text
 
 
 
 
 
 
25
 
26
  def get_response_from_chatbot(text):
27
  try:
@@ -147,18 +153,18 @@ with gr.Blocks(title='Talk to chatGPT') as demo:
147
  chatbot = gr.Chatbot(elem_id="chat_bot", visible=False).style(color_map=("green", "blue"))
148
  chatbot1 = gr.Chatbot(elem_id="chat_bot1").style(color_map=("green", "blue"))
149
  with gr.Row(elem_id="prompt_row"):
150
- #prompt_input_audio = gr.Audio(label = 'Record Audio Input',source="microphone",type="filepath")
151
  prompt_input = gr.Textbox(lines=2, label="Input text",show_label=True)
152
  chat_history = gr.Textbox(lines=4, label="prompt", visible=False)
153
- # transcribe_btn = gr.Button(value = "Transcribe").style(
154
- # margin=True,
155
- # rounded=(True, True, True, True),
156
- # width=100
157
- # )
158
- # transcribe_btn.click(fn=transcribe,
159
- # inputs=prompt_input_audio,
160
- # outputs=prompt_input
161
- # )
162
  submit_btn = gr.Button(value = "Submit",elem_id="submit-btn").style(
163
  margin=True,
164
  rounded=(True, True, True, True),
 
19
  device=device,
20
  )
21
 
22
+ all_special_ids = whisper_model.tokenizer.all_special_ids
23
+ transcribe_token_id = all_special_ids[-5]
24
+ translate_token_id = all_special_ids[-6]
25
+
26
+ def transcribe(audio):
27
+ task = "translate"
28
+ whisper_model.model.config.forced_decoder_ids = [[2, transcribe_token_id if task=="transcribe" else translate_token_id]]
29
+ text = whisper_model(audio)["text"]
30
+ return text
31
 
32
  def get_response_from_chatbot(text):
33
  try:
 
153
  chatbot = gr.Chatbot(elem_id="chat_bot", visible=False).style(color_map=("green", "blue"))
154
  chatbot1 = gr.Chatbot(elem_id="chat_bot1").style(color_map=("green", "blue"))
155
  with gr.Row(elem_id="prompt_row"):
156
+ prompt_input_audio = gr.Audio(label = 'Record Audio Input',source="microphone",type="filepath")
157
  prompt_input = gr.Textbox(lines=2, label="Input text",show_label=True)
158
  chat_history = gr.Textbox(lines=4, label="prompt", visible=False)
159
+ transcribe_btn = gr.Button(value = "Transcribe").style(
160
+ margin=True,
161
+ rounded=(True, True, True, True),
162
+ width=100
163
+ )
164
+ transcribe_btn.click(fn=transcribe,
165
+ inputs=prompt_input_audio,
166
+ outputs=prompt_input
167
+ )
168
  submit_btn = gr.Button(value = "Submit",elem_id="submit-btn").style(
169
  margin=True,
170
  rounded=(True, True, True, True),