paige-ai

/

Virchow2

like 45

Follow

Paige AI 49

Hi!

I am getting a runtime error when I try to execute the code recommended to use Virchow2 to generate embeddings. The code is the following:

from huggingface_hub import login

login()

import timm
import torch
from timm.data import resolve_data_config
from timm.data.transforms_factory import create_transform
from timm.layers import SwiGLUPacked
from PIL import Image
model = timm.create_model("hf-hub:paige-ai/Virchow2", pretrained=True, mlp_layer=SwiGLUPacked, act_layer=torch.nn.SiLU)

The error is the following:

RuntimeError: Error(s) in loading state_dict for VisionTransformer:
Unexpected key(s) in state_dict: "reg_token", "blocks.0.ls1.gamma", "blocks.0.ls2.gamma", "blocks.1.ls1.gamma", "blocks.1.ls2.gamma", "blocks.2.ls1.gamma", "blocks.2.ls2.gamma", "blocks.3.ls1.gamma", "blocks.3.ls2.gamma", "blocks.4.ls1.gamma", "blocks.4.ls2.gamma", "blocks.5.ls1.gamma", "blocks.5.ls2.gamma", "blocks.6.ls1.gamma", "blocks.6.ls2.gamma", "blocks.7.ls1.gamma", "blocks.7.ls2.gamma", "blocks.8.ls1.gamma", "blocks.8.ls2.gamma", "blocks.9.ls1.gamma", "blocks.9.ls2.gamma", "blocks.10.ls1.gamma", "blocks.10.ls2.gamma", "blocks.11.ls1.gamma", "blocks.11.ls2.gamma", "blocks.12.ls1.gamma", "blocks.12.ls2.gamma", "blocks.13.ls1.gamma", "blocks.13.ls2.gamma", "blocks.14.ls1.gamma", "blocks.14.ls2.gamma", "blocks.15.ls1.gamma", "blocks.15.ls2.gamma", "blocks.16.ls1.gamma", "blocks.16.ls2.gamma", "blocks.17.ls1.gamma", "blocks.17.ls2.gamma", "blocks.18.ls1.gamma", "blocks.18.ls2.gamma", "blocks.19.ls1.gamma", "blocks.19.ls2.gamma", "blocks.20.ls1.gamma", "blocks.20.ls2.gamma", "blocks.21.ls1.gamma", "blocks.21.ls2.gamma", "blocks.22.ls1.gamma", "blocks.22.ls2.gamma", "blocks.23.ls1.gamma", "blocks.23.ls2.gamma", "blocks.24.ls1.gamma", "blocks.24.ls2.gamma", "blocks.25.ls1.gamma", "blocks.25.ls2.gamma", "blocks.26.ls1.gamma", "blocks.26.ls2.gamma", "blocks.27.ls1.gamma", "blocks.27.ls2.gamma", "blocks.28.ls1.gamma", "blocks.28.ls2.gamma", "blocks.29.ls1.gamma", "blocks.29.ls2.gamma", "blocks.30.ls1.gamma", "blocks.30.ls2.gamma", "blocks.31.ls1.gamma", "blocks.31.ls2.gamma".
size mismatch for pos_embed: copying a param with shape torch.Size([1, 261, 1280]) from checkpoint, the shape in current model is torch.Size([1, 257, 1280]).

Any idea on how I can solve?

Thank you!