Create handler.py
Browse files- handler.py +70 -0
handler.py
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import torch
|
| 2 |
+
from transformers import AutoModelForCausalLM, AutoProcessor
|
| 3 |
+
from PIL import Image
|
| 4 |
+
import base64
|
| 5 |
+
import io
|
| 6 |
+
import logging
|
| 7 |
+
|
| 8 |
+
logger = logging.getLogger(__name__)
|
| 9 |
+
|
| 10 |
+
class EndpointHandler():
|
| 11 |
+
def __init__(self, path=""):
|
| 12 |
+
"""
|
| 13 |
+
ฟังก์ชันนี้จะทำงานแค่ครั้งเดียวตอนเริ่มต้น Endpoint เพื่อโหลดโมเดลรอไว้
|
| 14 |
+
"""
|
| 15 |
+
logger.info("Initializing model...")
|
| 16 |
+
self.device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 17 |
+
|
| 18 |
+
# โหลดโมเดลและ processor จาก path ที่ Hugging Face ส่งมาให้
|
| 19 |
+
self.model = AutoModelForCausalLM.from_pretrained(
|
| 20 |
+
path,
|
| 21 |
+
trust_remote_code=True,
|
| 22 |
+
torch_dtype=torch.bfloat16,
|
| 23 |
+
device_map=self.device,
|
| 24 |
+
# revision="b98e57b" # อาจจะต้องใช้ถ้ามีปัญหา
|
| 25 |
+
)
|
| 26 |
+
self.processor = AutoProcessor.from_pretrained(path, trust_remote_code=True)
|
| 27 |
+
logger.info("Model initialized successfully.")
|
| 28 |
+
|
| 29 |
+
def __call__(self, data):
|
| 30 |
+
"""
|
| 31 |
+
ฟังก์ชันนี้จะทำงานทุกครั้งที่มี request ส่งเข้ามาที่ API
|
| 32 |
+
"""
|
| 33 |
+
logger.info("Processing new request...")
|
| 34 |
+
# ดึงข้อมูลจาก request
|
| 35 |
+
inputs = data.pop("inputs", data)
|
| 36 |
+
image_b64 = inputs.get("image") # แนะนำให้ใช้ key ชื่อ "image"
|
| 37 |
+
|
| 38 |
+
if not image_b64:
|
| 39 |
+
return {"error": "Missing 'image' key with base64 encoded string in inputs."}
|
| 40 |
+
|
| 41 |
+
try:
|
| 42 |
+
# แปลง base64 string กลับเป็นรูปภาพ
|
| 43 |
+
image_bytes = base64.b64decode(image_b64)
|
| 44 |
+
image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
| 45 |
+
except Exception as e:
|
| 46 |
+
logger.error(f"Error decoding image: {e}")
|
| 47 |
+
return {"error": f"Invalid base64 image data. {e}"}
|
| 48 |
+
|
| 49 |
+
# สร้าง Prompt ตามรูปแบบที่โมเดลต้องการ
|
| 50 |
+
prompt = "<|user|>\n<image>\n<|assistant|>"
|
| 51 |
+
|
| 52 |
+
# เตรียมข้อมูลสำหรับส่งเข้าโมเดล
|
| 53 |
+
model_inputs = self.processor(text=prompt, images=image, return_tensors="pt").to(self.device, torch.bfloat16)
|
| 54 |
+
|
| 55 |
+
# รันโมเดลเพื่อสร้าง text
|
| 56 |
+
generated_ids = self.model.generate(
|
| 57 |
+
input_ids=model_inputs["input_ids"],
|
| 58 |
+
pixel_values=model_inputs["pixel_values"],
|
| 59 |
+
max_new_tokens=2048,
|
| 60 |
+
do_sample=False,
|
| 61 |
+
num_beams=1
|
| 62 |
+
)
|
| 63 |
+
|
| 64 |
+
# ถอดรหัสผลลัพธ์
|
| 65 |
+
generated_ids = generated_ids[:, model_inputs['input_ids'].shape[1]:]
|
| 66 |
+
response_text = self.processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
| 67 |
+
|
| 68 |
+
logger.info("Request processed successfully.")
|
| 69 |
+
# ส่งผลลัพธ์กลับในรูปแบบ JSON
|
| 70 |
+
return {"generated_text": response_text}
|