File size: 2,475 Bytes
7c7c31b | 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 |
import torch
import base64
import io
import re
from typing import Dict, List, Any
from PIL import Image
from transformers import AutoProcessor, AutoModelForImageTextToText
class EndpointHandler:
def __init__(self, path=""):
self.processor = AutoProcessor.from_pretrained(path, trust_remote_code=True)
self.model = AutoModelForImageTextToText.from_pretrained(
path,
torch_dtype=torch.float16,
device_map="auto",
trust_remote_code=True,
).eval()
self.device = next(self.model.parameters()).device
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
inputs_data = data.get("inputs", data)
# Accept base64 image
if isinstance(inputs_data, dict):
image_b64 = inputs_data.get("image", "")
prompt = inputs_data.get("prompt", "Text Recognition:")
elif isinstance(inputs_data, str):
image_b64 = inputs_data
prompt = "Text Recognition:"
else:
return [{"error": "Invalid input format"}]
# Decode image
try:
image_bytes = base64.b64decode(image_b64)
image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
except Exception as e:
return [{"error": f"Failed to decode image: {str(e)}"}]
# Build messages
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": image},
{"type": "text", "text": prompt},
],
}
]
# Process
text = self.processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
proc_inputs = self.processor(
text=[text], images=[image], padding=True, return_tensors="pt"
)
proc_inputs = {k: v.to(self.device) for k, v in proc_inputs.items()}
# Generate
with torch.no_grad():
output = self.model.generate(
**proc_inputs,
temperature=0.1,
max_new_tokens=8192,
do_sample=True,
)
prompt_len = proc_inputs["input_ids"].shape[1]
new_tokens = output[:, prompt_len:]
text_output = self.processor.tokenizer.batch_decode(
new_tokens, skip_special_tokens=True
)[0]
return [{"generated_text": text_output}]
|