inference.py

import io, base64
import numpy as np
from PIL import Image
import torch
from safetensors.torch import load_file
from model import TinyConv

LABELS = [str(i) for i in range(10)]

def load_model(path: str = ""):
    m = TinyConv().eval()
    try:
        state = load_file(f"{path}model.safetensors")
        m.load_state_dict(state, strict=True)
    except Exception:
        # Fallback to randomly initialized weights if file missing
        pass
    return m

MODEL = load_model()

def preprocess_pil(img: Image.Image) -> torch.Tensor:
    img = img.convert("L").resize((28, 28))
    x = torch.from_numpy(np.array(img)).float() / 255.0
    x = x.unsqueeze(0).unsqueeze(0)  # [1,1,28,28]
    return x

def predict_from_pil(img: Image.Image):
    x = preprocess_pil(img)
    with torch.inference_mode():
        logits = MODEL(x)
        probs = torch.softmax(logits, dim=-1)[0].tolist()
    return {label: float(probs[i]) for i, label in enumerate(LABELS)}

def predict_from_base64(b64: str):
    img = Image.open(io.BytesIO(base64.b64decode(b64)))
    return predict_from_pil(img)