Update app.py
Browse files
app.py
CHANGED
|
@@ -1,129 +1,8 @@
|
|
| 1 |
-
from fastapi import FastAPI, UploadFile, File
|
| 2 |
-
from transformers import pipeline, WhisperForConditionalGeneration, WhisperProcessor
|
| 3 |
import torch
|
| 4 |
-
import tempfile
|
| 5 |
-
import os
|
| 6 |
-
import time
|
| 7 |
-
from fastapi.responses import HTMLResponse
|
| 8 |
-
from fastapi.staticfiles import StaticFiles
|
| 9 |
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
print(f"CUDA device: {torch.cuda.get_device_name(torch.cuda.current_device())}")
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
# Load the Whisper model once during startup
|
| 18 |
-
device = 0 if torch.cuda.is_available() else -1 # Use GPU if available, otherwise CPU
|
| 19 |
-
asr_pipeline = pipeline(model="openai/whisper-large", device=device) # Initialize Whisper model
|
| 20 |
-
# asr_pipeline = pipeline( model="openai/whisper-small", device=device, language="pt")
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
# Basic GET endpoint
|
| 24 |
-
@app.get("/")
|
| 25 |
-
def read_root():
|
| 26 |
-
return {"message": "Welcome to the FastAPI app on Hugging Face Spaces!"}
|
| 27 |
-
|
| 28 |
-
# POST endpoint to transcribe audio
|
| 29 |
-
@app.post("/transcribe/")
|
| 30 |
-
async def transcribe_audio(file: UploadFile = File(...)):
|
| 31 |
-
start_time = time.time()
|
| 32 |
-
|
| 33 |
-
# Save the uploaded file using a temporary file manager
|
| 34 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio_file:
|
| 35 |
-
temp_audio_file.write(await file.read())
|
| 36 |
-
temp_file_path = temp_audio_file.name
|
| 37 |
-
|
| 38 |
-
# Transcribe the audio with long-form generation enabled
|
| 39 |
-
transcription_start = time.time()
|
| 40 |
-
transcription = asr_pipeline(temp_file_path, return_timestamps=True) # Enable timestamp return for long audio files
|
| 41 |
-
transcription_end = time.time()
|
| 42 |
-
|
| 43 |
-
# Clean up temporary file after use
|
| 44 |
-
os.remove(temp_file_path)
|
| 45 |
-
|
| 46 |
-
# Log time durations
|
| 47 |
-
end_time = time.time()
|
| 48 |
-
print(f"Time to transcribe audio: {transcription_end - transcription_start:.4f} seconds")
|
| 49 |
-
print(f"Total execution time: {end_time - start_time:.4f} seconds")
|
| 50 |
-
|
| 51 |
-
return {"transcription": transcription['text']}
|
| 52 |
-
|
| 53 |
-
@app.get("/playground/", response_class=HTMLResponse)
|
| 54 |
-
def playground():
|
| 55 |
-
html_content = """
|
| 56 |
-
<!DOCTYPE html>
|
| 57 |
-
<html lang="en">
|
| 58 |
-
<head>
|
| 59 |
-
<meta charset="UTF-8">
|
| 60 |
-
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
| 61 |
-
<title>Voice Recorder</title>
|
| 62 |
-
</head>
|
| 63 |
-
<body>
|
| 64 |
-
<h1>Record your voice</h1>
|
| 65 |
-
<button id="startBtn">Start Recording</button>
|
| 66 |
-
<button id="stopBtn" disabled>Stop Recording</button>
|
| 67 |
-
<p id="status">Press start to record your voice...</p>
|
| 68 |
-
|
| 69 |
-
<audio id="audioPlayback" controls style="display:none;"></audio>
|
| 70 |
-
<script>
|
| 71 |
-
let mediaRecorder;
|
| 72 |
-
let audioChunks = [];
|
| 73 |
-
|
| 74 |
-
const startBtn = document.getElementById('startBtn');
|
| 75 |
-
const stopBtn = document.getElementById('stopBtn');
|
| 76 |
-
const status = document.getElementById('status');
|
| 77 |
-
const audioPlayback = document.getElementById('audioPlayback');
|
| 78 |
-
|
| 79 |
-
// Start Recording
|
| 80 |
-
startBtn.addEventListener('click', async () => {
|
| 81 |
-
const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
|
| 82 |
-
mediaRecorder = new MediaRecorder(stream);
|
| 83 |
-
mediaRecorder.start();
|
| 84 |
-
|
| 85 |
-
status.textContent = 'Recording...';
|
| 86 |
-
startBtn.disabled = true;
|
| 87 |
-
stopBtn.disabled = false;
|
| 88 |
-
|
| 89 |
-
mediaRecorder.ondataavailable = event => {
|
| 90 |
-
audioChunks.push(event.data);
|
| 91 |
-
};
|
| 92 |
-
});
|
| 93 |
-
|
| 94 |
-
// Stop Recording
|
| 95 |
-
stopBtn.addEventListener('click', () => {
|
| 96 |
-
mediaRecorder.stop();
|
| 97 |
-
mediaRecorder.onstop = async () => {
|
| 98 |
-
status.textContent = 'Recording stopped. Preparing to send...';
|
| 99 |
-
const audioBlob = new Blob(audioChunks, { type: 'audio/wav' });
|
| 100 |
-
const audioUrl = URL.createObjectURL(audioBlob);
|
| 101 |
-
audioPlayback.src = audioUrl;
|
| 102 |
-
audioPlayback.style.display = 'block';
|
| 103 |
-
audioChunks = [];
|
| 104 |
-
|
| 105 |
-
// Send audio blob to FastAPI endpoint
|
| 106 |
-
const formData = new FormData();
|
| 107 |
-
formData.append('file', audioBlob, 'recording.wav');
|
| 108 |
-
|
| 109 |
-
const response = await fetch('/transcribe/', {
|
| 110 |
-
method: 'POST',
|
| 111 |
-
body: formData,
|
| 112 |
-
});
|
| 113 |
-
|
| 114 |
-
const result = await response.json();
|
| 115 |
-
status.textContent = 'Transcription: ' + result.transcription;
|
| 116 |
-
};
|
| 117 |
-
|
| 118 |
-
startBtn.disabled = false;
|
| 119 |
-
stopBtn.disabled = true;
|
| 120 |
-
});
|
| 121 |
-
</script>
|
| 122 |
-
</body>
|
| 123 |
-
</html>
|
| 124 |
-
"""
|
| 125 |
-
return HTMLResponse(content=html_content)
|
| 126 |
-
# If running as the main module, start Uvicorn
|
| 127 |
-
if __name__ == "__main__":
|
| 128 |
-
import uvicorn
|
| 129 |
-
uvicorn.run(app, host="0.0.0.0", port=7860)
|
|
|
|
|
|
|
|
|
|
| 1 |
import torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
+
if torch.cuda.is_available():
|
| 4 |
+
device = torch.device("cuda")
|
| 5 |
+
print(f"CUDA device: {torch.cuda.get_device_name(torch.cuda.current_device())}")
|
| 6 |
+
else:
|
| 7 |
+
device = torch.device("cpu")
|
| 8 |
+
print("CUDA is not available, using CPU.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|