File size: 4,174 Bytes
253f083 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 |
from flask import Flask, request, jsonify, Response # Add Response here
from faster_whisper import WhisperModel
import torch
import io
import time
import datetime
from threading import Semaphore
import os
from werkzeug.utils import secure_filename
import tempfile
app = Flask(__name__)
# Configuration
MAX_CONCURRENT_REQUESTS = 2 # Adjust based on your server capacity
MAX_AUDIO_DURATION = 60 * 30 # 30 minutes maximum audio duration (adjust as needed)
TEMPORARY_FOLDER = tempfile.gettempdir()
ALLOWED_EXTENSIONS = {'mp3', 'wav', 'ogg', 'm4a', 'flac'}
# Device check for faster-whisper
device = "cuda" if torch.cuda.is_available() else "cpu"
compute_type = "float16" if device == "cuda" else "int8"
print(f"Using device: {device} with compute_type: {compute_type}")
# Faster Whisper setup with optimized parameters for long audio
beamsize = 2 # Slightly larger beam size can help with long-form accuracy
wmodel = WhisperModel(
"guillaumekln/faster-whisper-small",
device=device,
compute_type=compute_type,
download_root="./model_cache" # Cache model to avoid re-downloading
)
# Concurrency control
request_semaphore = Semaphore(MAX_CONCURRENT_REQUESTS)
active_requests = 0
def allowed_file(filename):
return '.' in filename and \
filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
def cleanup_temp_files(file_path):
"""Ensure temporary files are deleted after processing"""
try:
if os.path.exists(file_path):
os.remove(file_path)
except Exception as e:
print(f"Error cleaning up temp file {file_path}: {str(e)}")
@app.route("/health", methods=["GET"])
def health_check():
"""Endpoint to check if API is running"""
return jsonify({
'status': 'API is running',
'timestamp': datetime.datetime.now().isoformat(),
'device': device,
'compute_type': compute_type,
'active_requests': active_requests,
'max_duration_supported': MAX_AUDIO_DURATION
})
@app.route("/status/busy", methods=["GET"])
def server_busy():
"""Endpoint to check if server is busy"""
is_busy = active_requests >= MAX_CONCURRENT_REQUESTS
return jsonify({
'is_busy': is_busy,
'active_requests': active_requests,
'max_capacity': MAX_CONCURRENT_REQUESTS
})
@app.route("/whisper_transcribe", methods=["POST"])
def whisper_transcribe():
global active_requests
if not request_semaphore.acquire(blocking=False):
return jsonify({'error': 'Server busy'}), 503
active_requests += 1
start_time = time.time()
temp_file_path = None
try:
if 'audio' not in request.files:
return jsonify({'error': 'No file provided'}), 400
audio_file = request.files['audio']
if not (audio_file and allowed_file(audio_file.filename)):
return jsonify({'error': 'Invalid file format'}), 400
temp_file_path = os.path.join(TEMPORARY_FOLDER, secure_filename(audio_file.filename))
audio_file.save(temp_file_path)
segments, _ = wmodel.transcribe(
temp_file_path,
beam_size=beamsize,
vad_filter=True,
without_timestamps=True, # Ensure timestamps are not included
compression_ratio_threshold=2.4,
word_timestamps=False
)
full_text = " ".join(segment.text for segment in segments)
return jsonify({'transcription': full_text}), 200
except Exception as e:
return jsonify({'error': str(e)}), 500
finally:
if temp_file_path:
cleanup_temp_files(temp_file_path)
active_requests -= 1
request_semaphore.release()
print(f"Processed in {time.time()-start_time:.2f}s (Active: {active_requests})")
if __name__ == "__main__":
# Create temporary folder if it doesn't exist
if not os.path.exists(TEMPORARY_FOLDER):
os.makedirs(TEMPORARY_FOLDER)
app.run(host="0.0.0.0", port=7860, threaded=True) |