minimal_script.py

import os
import torch
import numpy as np
import torch.nn as nn
import lightning.pytorch as pl
from safetensors.torch import save_file, load_file
from transformers import AutoImageProcessor, AutoModel
from transformers.image_utils import load_image

class EmbeddingNetwork(nn.Module):
    def __init__(self):
        super(EmbeddingNetwork, self).__init__()
        self.fc1 = nn.Linear(1280, 256)
        self.dropout1 = nn.Dropout(0.33)
        self.fc2 = nn.Linear(256, 128)
        self.dropout2 = nn.Dropout(0.33)
        self.fc3 = nn.Linear(128, 7)
        self.act = nn.ReLU(inplace=True)

    def forward(self, x):
        x = self.fc1(x)
        #x = self.dropout1(x)
        x = self.act(x)
        x = self.fc2(x)
        #x = self.dropout2(x)
        x = self.act(x)
        x = self.fc3(x)
        return x



class PLModule(pl.LightningModule):
    def __init__(self):
        super().__init__()
        self.save_hyperparameters()
        self.network = EmbeddingNetwork()

    def forward(self, x):
        return self.network(x)

    def predict_step(self, batch, batch_idx, dataloader_idx=0):
        outputs = self.forward(batch[0])
        return outputs, batch[1]



if __name__ == '__main__':
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    
    embd_model = EmbeddingNetwork().to(device=device, dtype=torch.bfloat16)
    state_dict = load_file("Style Embedder v4.safetensors")
    embd_model.load_state_dict(state_dict)

    token = 'Enter your huggingface token here'
    processor = AutoImageProcessor.from_pretrained("facebook/dinov3-vits16-pretrain-lvd1689m",
                                                   do_resize=False, token=token)
    dino_model = AutoModel.from_pretrained("facebook/dinov3-vith16plus-pretrain-lvd1689m", token=token, device_map="auto",
                                           dtype=torch.bfloat16)
    image = load_image('images_for_style_embedding/6857740.webp')
    input = processor(images=image, return_tensors="pt").to(device=dino_model.device, dtype=torch.bfloat16)
    output = dino_model(**input)
    last_hidden_states = output.last_hidden_state
    cls_token = last_hidden_states[:, 0, :]

    pred = embd_model(cls_token).cpu()
    print(pred)