Spaces:
Runtime error
Runtime error
Upload 3 files
Browse files
app.py
CHANGED
|
@@ -3,12 +3,26 @@ import gradio as gr
|
|
| 3 |
|
| 4 |
fn.load_model()
|
| 5 |
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
if __name__ == '__main__':
|
| 14 |
demo.launch()
|
|
|
|
| 3 |
|
| 4 |
fn.load_model()
|
| 5 |
|
| 6 |
+
with gr.Blocks() as demo:
|
| 7 |
+
audio = gr.Audio(sources="upload", type="filepath")
|
| 8 |
+
model = gr.Dropdown(value='large-v3', choices=["tiny", "base", "small", "medium", "large", "large-v2", "large-v3"])
|
| 9 |
+
run_button = gr.Button(value='Run')
|
| 10 |
+
prompt = gr.Textbox(label='prompt')
|
| 11 |
+
set_button = gr.Button(value='Set Prompt')
|
| 12 |
+
text_only = gr.Textbox(label='output')
|
| 13 |
+
text_with_timestamps = gr.Textbox(label='timestamps')
|
| 14 |
+
|
| 15 |
+
run_button.click(
|
| 16 |
+
fn=fn.speech_to_text,
|
| 17 |
+
inputs=[audio, model],
|
| 18 |
+
outputs=[text_only, text_with_timestamps],
|
| 19 |
+
)
|
| 20 |
+
|
| 21 |
+
set_button.click(
|
| 22 |
+
fn=fn.set_prompt,
|
| 23 |
+
inputs=[prompt],
|
| 24 |
+
outputs=[],
|
| 25 |
+
)
|
| 26 |
|
| 27 |
if __name__ == '__main__':
|
| 28 |
demo.launch()
|
fn.py
CHANGED
|
@@ -10,6 +10,7 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
|
| 10 |
|
| 11 |
model = None
|
| 12 |
pipe = None
|
|
|
|
| 13 |
|
| 14 |
def load_model():
|
| 15 |
global model, pipe
|
|
@@ -28,14 +29,21 @@ def load_model():
|
|
| 28 |
device=device,
|
| 29 |
)
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
def speech_to_text(audio_file, _model_size = None):
|
| 32 |
-
global model, pipe
|
| 33 |
|
| 34 |
if not model:
|
| 35 |
load_model()
|
| 36 |
|
| 37 |
# run inference
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
try:
|
| 41 |
res = json.dumps(result)
|
|
|
|
| 10 |
|
| 11 |
model = None
|
| 12 |
pipe = None
|
| 13 |
+
initial_prompt = None
|
| 14 |
|
| 15 |
def load_model():
|
| 16 |
global model, pipe
|
|
|
|
| 29 |
device=device,
|
| 30 |
)
|
| 31 |
|
| 32 |
+
def set_prompt(prompt):
|
| 33 |
+
global initial_prompt
|
| 34 |
+
initial_prompt = prompt
|
| 35 |
+
|
| 36 |
def speech_to_text(audio_file, _model_size = None):
|
| 37 |
+
global model, pipe, initial_prompt
|
| 38 |
|
| 39 |
if not model:
|
| 40 |
load_model()
|
| 41 |
|
| 42 |
# run inference
|
| 43 |
+
generate_kwargs = {}
|
| 44 |
+
if initial_prompt:
|
| 45 |
+
generate_kwargs['prompt_ids'] = pipe.tokenizer.get_prompt_ids(prompt, return_tensors="pt").to(device)
|
| 46 |
+
result = pipe(audio_file, generate_kwargs=generate_kwargs)
|
| 47 |
|
| 48 |
try:
|
| 49 |
res = json.dumps(result)
|
main.py
CHANGED
|
@@ -40,3 +40,12 @@ async def transcribe_audio(file: UploadFile = Form(...)):
|
|
| 40 |
return {"transcription": text_only, "text_with_timestamps": text_with_timestamps}
|
| 41 |
except Exception as e:
|
| 42 |
return {"error": str(e)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
return {"transcription": text_only, "text_with_timestamps": text_with_timestamps}
|
| 41 |
except Exception as e:
|
| 42 |
return {"error": str(e)}
|
| 43 |
+
|
| 44 |
+
@app.post("/set_prompt")
|
| 45 |
+
async def set_prompt(prompt: str):
|
| 46 |
+
try:
|
| 47 |
+
fn.set_prompt(prompt)
|
| 48 |
+
|
| 49 |
+
return {"status": 0}
|
| 50 |
+
except Exception as e:
|
| 51 |
+
return {"error": str(e)}
|