Spaces:
Paused
Paused
File size: 9,251 Bytes
68f4b63 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 | import torch
from fastapi import FastAPI, HTTPException, Request
from fastapi.responses import HTMLResponse
from fastapi.staticfiles import StaticFiles
from pydantic import BaseModel
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
from IndicTransToolkit.processor import IndicProcessor
import os
import traceback
app = FastAPI()
# Configuration
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Running on device: {DEVICE}")
# Load models
MODELS = {}
HF_TOKEN = os.environ.get("HF_TOKEN")
def load_model(name, repo_id):
print(f"Loading {name} from {repo_id}...")
try:
tokenizer = AutoTokenizer.from_pretrained(
repo_id,
trust_remote_code=True,
token=HF_TOKEN
)
# Removed flash_attention_2 to fix 'NoneType' shape error on T4
model = AutoModelForSeq2SeqLM.from_pretrained(
repo_id,
trust_remote_code=True,
torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
token=HF_TOKEN
).to(DEVICE)
model.eval()
return {"tokenizer": tokenizer, "model": model}
except Exception as e:
print(f"Failed to load {name}: {e}")
raise e
# Load on startup
@app.on_event("startup")
async def startup_event():
global MODELS, ip
if not HF_TOKEN:
print("WARNING: HF_TOKEN environment variable is not set. Gated models may fail to load.")
# 1. English to Indic
MODELS["en-indic"] = load_model("en-indic", "ai4bharat/indictrans2-en-indic-dist-200M")
# 2. Indic to English
MODELS["indic-en"] = load_model("indic-en", "ai4bharat/indictrans2-indic-en-dist-200M")
# Processor
ip = IndicProcessor(inference=True)
print("All models loaded successfully.")
class TranslationRequest(BaseModel):
text: str
source_lang: str
target_lang: str
@app.post("/translate")
async def translate(request: TranslationRequest):
try:
src = request.source_lang
tgt = request.target_lang
text = request.text
if not text:
return {"translated_text": ""}
if src.startswith("eng"):
model_key = "en-indic"
elif tgt.startswith("eng"):
model_key = "indic-en"
else:
raise HTTPException(status_code=400, detail="Direct Indic-to-Indic translation not supported.")
if model_key not in MODELS:
raise HTTPException(status_code=500, detail=f"Model {model_key} failed to load on startup.")
print(f"Translating {model_key}: {src} -> {tgt} (len: {len(text)})")
bundle = MODELS[model_key]
tokenizer = bundle["tokenizer"]
model = bundle["model"]
# Preprocess
batch = ip.preprocess_batch([text], src_lang=src, tgt_lang=tgt)
# Tokenize
inputs = tokenizer(
batch,
truncation=True,
padding="longest",
return_tensors="pt",
return_attention_mask=True,
).to(DEVICE)
# Generate
with torch.no_grad():
generated_tokens = model.generate(
**inputs,
use_cache=False,
min_length=0,
max_length=2048,
num_beams=5,
num_return_sequences=1,
)
# Decode
decoded_tokens = tokenizer.batch_decode(
generated_tokens,
skip_special_tokens=True,
clean_up_tokenization_spaces=True,
)
# Postprocess
translations = ip.postprocess_batch(decoded_tokens, lang=tgt)
return {"translated_text": translations[0]}
except Exception as e:
traceback.print_exc()
print(f"Error: {e}")
raise HTTPException(status_code=500, detail=str(e))
@app.get("/", response_class=HTMLResponse)
def read_root():
return """
<!DOCTYPE html>
<html>
<head>
<title>Noa AI Translator</title>
<script src="https://cdn.tailwindcss.com"></script>
</head>
<body class="bg-gray-50 min-h-screen p-8">
<div class="max-w-2xl mx-auto bg-white rounded-xl shadow-md p-6">
<h1 class="text-2xl font-bold mb-6 text-gray-800">Noa AI Translator</h1>
<div class="space-y-4">
<div class="grid grid-cols-2 gap-4">
<div>
<label class="block text-sm font-medium text-gray-700 mb-1">Source Language</label>
<select id="sourceLang" class="w-full border rounded-md p-2">
<option value="eng_Latn">English</option>
<option value="hin_Deva">Hindi</option>
<option value="tam_Taml">Tamil</option>
<option value="tel_Telu">Telugu</option>
<option value="kan_Knda">Kannada</option>
<option value="mal_Mlym">Malayalam</option>
<option value="mar_Deva">Marathi</option>
<option value="guj_Gujr">Gujarati</option>
<option value="ben_Beng">Bengali</option>
<option value="asm_Beng">Assamese</option>
<option value="pan_Guru">Punjabi</option>
</select>
</div>
<div>
<label class="block text-sm font-medium text-gray-700 mb-1">Target Language</label>
<select id="targetLang" class="w-full border rounded-md p-2">
<option value="hin_Deva">Hindi</option>
<option value="eng_Latn">English</option>
<option value="tam_Taml">Tamil</option>
<option value="tel_Telu">Telugu</option>
<option value="kan_Knda">Kannada</option>
<option value="mal_Mlym">Malayalam</option>
<option value="mar_Deva">Marathi</option>
<option value="guj_Gujr">Gujarati</option>
<option value="ben_Beng">Bengali</option>
<option value="asm_Beng">Assamese</option>
<option value="pan_Guru">Punjabi</option>
</select>
</div>
</div>
<div>
<label class="block text-sm font-medium text-gray-700 mb-1">Input Text</label>
<textarea id="inputText" rows="6" class="w-full border rounded-md p-2" placeholder="Enter text here..."></textarea>
</div>
<button onclick="translateText()" id="translateBtn" class="w-full bg-blue-600 text-white py-2 px-4 rounded-md hover:bg-blue-700 transition-colors font-medium">
Translate
</button>
<div>
<label class="block text-sm font-medium text-gray-700 mb-1">Translation</label>
<div id="outputText" class="w-full border rounded-md p-4 min-h-[150px] bg-gray-50 whitespace-pre-wrap"></div>
</div>
</div>
</div>
<script>
async function translateText() {
const btn = document.getElementById('translateBtn');
const output = document.getElementById('outputText');
const text = document.getElementById('inputText').value;
const sourceLang = document.getElementById('sourceLang').value;
const targetLang = document.getElementById('targetLang').value;
if (!text) return;
btn.disabled = true;
btn.textContent = 'Translating...';
output.textContent = '';
try {
const response = await fetch('/translate', {
method: 'POST',
headers: {
'Content-Type': 'application/json',
},
body: JSON.stringify({
text: text,
source_lang: sourceLang,
target_lang: targetLang
})
});
const data = await response.json();
if (response.ok) {
output.textContent = data.translated_text;
output.classList.remove('text-red-500');
} else {
output.textContent = 'Error: ' + (data.detail || 'Translation failed');
output.classList.add('text-red-500');
}
} catch (e) {
output.textContent = 'Error: ' + e.message;
output.classList.add('text-red-500');
} finally {
btn.disabled = false;
btn.textContent = 'Translate';
}
}
</script>
</body>
</html>
"""
|