[BUG] timm is unable to load the huggingface custom model

[fangyynb]

Describe the bug

import timm

model_list = timm.list_models()
print(model_list)

The above is my code. The following error was reported after execution:

Traceback (most recent call last):
File "d:\Programmer\important things\lsnet-master\fy-test.py", line 4, in
model = timm.create_model(
File "C:\Users\fangy\miniconda3\envs\gcdp\lib\site-packages\timm\models_factory.py", line 134, in create_model
raise RuntimeError('Unknown model (%s)' % model_name)
RuntimeError: Unknown model (lsnet_t)

To Reproduce
Steps to reproduce the behavior:

1. install python=3.8
   2.install timm=1.0.17 and huggingface-hub=0.33.2
   3.then run the program

Expected behavior

lsnet_t should be created correctly

Screenshots

Desktop (please complete the following information):

• OS: windows11
• timm=1.0.17 and huggingface-hub=0.33.2 python=3.8
• torch=2.4.1+cu124

[rwightman]

@fangyynb I really can't tell what's going on here, but if the model requires an implementation that isn't built into timm, it has to have model entrypoint functions adhering to timm API that are registered with the @register_model dectorator, and whatever module has those has to be imported into the python environment before creating the model, whether the weights are on the hub or not

EDIT: I looked closer, found their model, you need to checkout their codebase and use relative to their, they didn't create a package (the lsnet below is not real) or adhere to the registration API closely enough to do something like the snippet ... ideally if they fixed their code you'd be able to do

pip install timm lsnet

import timm
import lsnet  # <-- this is needed to register custom models but not in the example, so you have to run within their repo path
import torch
from PIL import Image
import requests
from timm.data import resolve_data_config, create_transform

# Load the model
model = timm.create_model(
    'hf_hub:jameslahm/lsnet_b',
    pretrained=True
)
model.eval()

# Load and transform image
# Example using a URL:
url = 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
img = Image.open(requests.get(url, stream=True).raw)

config = resolve_data_config({}, model=model)
transform = create_transform(**config)
input_tensor = transform(img).unsqueeze(0) # transform and add batch dimension

# Make prediction
with torch.no_grad():
    output = model(input_tensor)
probabilities = torch.nn.functional.softmax(output[0], dim=0)

# Get top 5 predictions
top5_prob, top5_catid = torch.topk(probabilities, 5)
# Assuming you have imagenet labels list 'imagenet_labels'
# for i in range(top5_prob.size(0)):
#     print(imagenet_labels[top5_catid[i]], top5_prob[i].item())

[rwightman]

There is also at least one obvious mistake, they use hf_hub instead of hf_hub_id for the pretrained cfg registration. So I'd file issues there and reference this.

[fangyynb]

@rwightman Thank you so much for your reply! I'll look forward to the fix.