AdityaAjithKumar commited on
Commit
9c6f85e
·
1 Parent(s): af67663

fixed the API

Browse files
Files changed (1) hide show
  1. app.py +10 -6
app.py CHANGED
@@ -47,7 +47,7 @@ async def ocr(image: UploadFile = File(...), prompt: str = Form("Extract all tex
47
  logger.info(f"📸 OCR request received | Filename: {image.filename} | Prompt: {prompt[:100]}...")
48
 
49
  try:
50
- img = Image.open(io.BytesIO(await image.read()))
51
 
52
  messages = [{"role": "user", "content": [
53
  {"type": "image", "image": img},
@@ -55,10 +55,16 @@ async def ocr(image: UploadFile = File(...), prompt: str = Form("Extract all tex
55
  ]}]
56
 
57
  text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
58
- inputs = processor(text, images=img, return_tensors="pt").to("cpu")
59
 
60
- outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.3, do_sample=False)
61
- response = processor.decode(outputs[0], skip_special_tokens=True)
 
 
 
 
 
 
 
62
 
63
  logger.info("✅ OCR completed successfully")
64
  return {"extracted_text": response}
@@ -66,8 +72,6 @@ async def ocr(image: UploadFile = File(...), prompt: str = Form("Extract all tex
66
  except Exception as e:
67
  logger.error(f"❌ OCR error: {e}")
68
  return {"error": str(e)}
69
-
70
-
71
  # === ROOT ===
72
  @app.get("/")
73
  async def root():
 
47
  logger.info(f"📸 OCR request received | Filename: {image.filename} | Prompt: {prompt[:100]}...")
48
 
49
  try:
50
+ img = Image.open(io.BytesIO(await image.read())).convert("RGB") # ensure RGB
51
 
52
  messages = [{"role": "user", "content": [
53
  {"type": "image", "image": img},
 
55
  ]}]
56
 
57
  text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
 
58
 
59
+ # ✅ Fix: use keyword `text=` to avoid positional conflict with `images=`
60
+ inputs = processor(text=text, images=[img], return_tensors="pt").to("cpu")
61
+
62
+ outputs = model.generate(**inputs, max_new_tokens=512,
63
+ do_sample=False) # removed temperature when do_sample=False
64
+
65
+ # ✅ Fix: slice only the newly generated tokens to avoid echoing the prompt
66
+ input_len = inputs["input_ids"].shape[1]
67
+ response = processor.decode(outputs[0][input_len:], skip_special_tokens=True)
68
 
69
  logger.info("✅ OCR completed successfully")
70
  return {"extracted_text": response}
 
72
  except Exception as e:
73
  logger.error(f"❌ OCR error: {e}")
74
  return {"error": str(e)}
 
 
75
  # === ROOT ===
76
  @app.get("/")
77
  async def root():