Spaces:
Runtime error
Runtime error
File size: 5,733 Bytes
80cb407 1e22d83 80cb407 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 |
from flask import Flask, render_template, request
import whisper
import tempfile
import os
import time
import torch
import numpy as np
import requests
from tqdm import tqdm
from transformers import BertTokenizer
from model.multi_class_model import MultiClassModel # Adjust if needed
import lightning as L
app = Flask(__name__)
# === CONFIG ===
CHECKPOINT_URL = "https://huggingface.co/nenafem/original_split_synthesized/resolve/main/original_split_synthesized.ckpt?download=true"
CHECKPOINT_PATH = "final_checkpoint/original_split_synthesized.ckpt"
AGE_LABELS = ["semua usia", "anak", "remaja", "dewasa"]
# === FUNCTION TO DOWNLOAD CKPT IF NEEDED ===
def download_checkpoint_if_needed(url, save_path):
if not os.path.exists(save_path):
os.makedirs(os.path.dirname(save_path), exist_ok=True)
print(f"📥 Downloading model checkpoint from {url}...")
response = requests.get(url, stream=True, timeout=10)
if response.status_code == 200:
total = int(response.headers.get("content-length", 0))
with open(save_path, 'wb') as f, tqdm(total=total, unit='B', unit_scale=True, desc="Downloading") as pbar:
for chunk in response.iter_content(1024):
f.write(chunk)
pbar.update(len(chunk))
print("✅ Checkpoint downloaded!")
else:
raise Exception(f"❌ Failed to download: {response.status_code}")
# === INITIAL SETUP: Download & Load Model ===
download_checkpoint_if_needed(CHECKPOINT_URL, CHECKPOINT_PATH)
# Load tokenizer
tokenizer = BertTokenizer.from_pretrained('indolem/indobert-base-uncased')
# Load model from checkpoint
model = MultiClassModel.load_from_checkpoint(
CHECKPOINT_PATH,
n_out=4,
dropout=0.3,
lr=1e-5
)
model.eval()
# === ROUTES ===
@app.route('/', methods=['GET'])
def index():
return render_template('index.html')
@app.route('/transcribe', methods=['POST'])
def transcribe():
try:
# Load Whisper with Indonesian language support (large / turbo)
# https://github.com/openai/whisper
whisper_model = whisper.load_model("large")
# Start measuring time
start_time = time.time()
audio_file = request.files['file']
if audio_file:
# Save uploaded audio to temp file
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
temp_audio.write(audio_file.read())
temp_audio_path = temp_audio.name
# Step 1: Transcribe
transcription = whisper_model.transcribe(temp_audio_path, language="id")
os.remove(temp_audio_path)
transcribed_text = transcription["text"]
# Step 2: BERT Prediction
encoding = tokenizer.encode_plus(
transcribed_text,
add_special_tokens=True,
max_length=512,
return_token_type_ids=True,
padding="max_length",
return_attention_mask=True,
return_tensors='pt',
)
with torch.no_grad():
prediction = model(
encoding["input_ids"],
encoding["attention_mask"],
encoding["token_type_ids"]
)
logits = prediction
probabilities = torch.nn.functional.softmax(logits, dim=1).cpu().numpy().flatten()
predicted_class = np.argmax(probabilities)
predicted_label = AGE_LABELS[predicted_class]
prob_results = [(label, f"{prob:.4f}") for label, prob in zip(AGE_LABELS, probabilities)]
# Stop timer
end_time = time.time()
total_time = end_time - start_time
formatted_time = f"{total_time:.2f} seconds"
return render_template(
'transcribe.html',
task=transcribed_text,
prediction=predicted_label,
probabilities=prob_results,
total_time=formatted_time
)
except Exception as e:
print("Error:", e)
return str(e)
@app.route('/predict-text', methods=['POST'])
def predict_text():
try:
user_lyrics = request.form.get('lyrics', '').strip()
if not user_lyrics:
return "No lyrics provided.", 400
# Start timer
start_time = time.time()
encoding = tokenizer.encode_plus(
user_lyrics,
add_special_tokens=True,
max_length=512,
return_token_type_ids=True,
padding="max_length",
return_attention_mask=True,
return_tensors='pt',
)
with torch.no_grad():
prediction = model(
encoding["input_ids"],
encoding["attention_mask"],
encoding["token_type_ids"]
)
logits = prediction
probabilities = torch.nn.functional.softmax(logits, dim=1).cpu().numpy().flatten()
predicted_class = np.argmax(probabilities)
predicted_label = AGE_LABELS[predicted_class]
prob_results = [(label, f"{prob:.4f}") for label, prob in zip(AGE_LABELS, probabilities)]
# End timer
end_time = time.time()
total_time = f"{end_time - start_time:.2f} seconds"
return render_template(
'transcribe.html',
task=user_lyrics,
prediction=predicted_label,
probabilities=prob_results,
total_time=total_time
)
except Exception as e:
print("❌ Error in predict-text:", e)
return str(e), 500
if __name__ == "__main__":
app.run(debug=True)
|