File size: 1,530 Bytes
1c42c71 |
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 |
import torch
from pathlib import Path
from timm.models.vision_transformer import VisionTransformer
from functools import partial
from torch import nn
from huggingface_hub import snapshot_download
def load_model_from_huggingface(repo_id, model_filename):
# Download model files
model_path = snapshot_download(repo_id=repo_id, revision="main")
model_weights_path = Path(model_path) / model_filename
# Load model weights
model_weights = torch.load(model_weights_path)['model']
# Define the model (ensure this matches your model's architecture)
model = VisionTransformer(patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6)).eval()
# Load the weights into the model
loading = model.load_state_dict(model_weights, strict=False)
return model, loading
def process_single_image(image_path, input_size=224, dataset_mean=[0.3464, 0.2280, 0.2228], dataset_std=[0.2520, 0.2128, 0.2093]):
# Define the transformations
transform = T.Compose([
T.Resize((input_size, input_size)),
T.ToTensor(),
T.Normalize(mean=dataset_mean, std=dataset_std)
])
# Open the image
image = Image.open(image_path).convert('RGB')
# Apply the transformations
processed_image = transform(image)
return processed_image
device = "cuda"
dtype = torch.float16
model, loading_info = load_model_from_huggingface("egeozsoy/EndoViT", "endovit.pth")
model = model.to(device, dtype)
print(loading_info) |