pytorch-image-models

pytorch-image-models copied to clipboard

Follow recommendations from [PEP 561 – Distributing and Packaging Type Information](https://peps.python.org/pep-0561/).

antoinebrl

Hey, as you have already noticed based on many of your articles, there is basically zero reproducibility of your results based on the network models only. The accuracies very much...

svobora

### Model description The Kolmogorov–Arnold Transformer (KAT) replaces the standard MLP layers in transformers with Kolmogorov-Arnold Network (KAN) layers, improving the model's expressiveness and overall performance. ### Open source status...

Adamdad

https://github.com/altair199797/LowFormer Also pinging @altair199797

ogencoglu

**Is your feature request related to a problem? Please describe.** Model request: LW-DETR https://github.com/Atten4Vis/LW-DETR **Additional context** It would be very useful to benchmark this against mobilenet v4

Immortalin

[FEATURE] Could you introduce abstract classes or interfaces for clarity and easier extension?

1

**Is your feature request related to a problem? Please describe.** First of all thanks for creating and maintaining this great library. We're also using it to accelerate quick experiments. However,...

cgebbe

Add initial weights for my first 3 naflexvit_base models

1

Testing different pos embed configs. Fix small tracing issue for 2d learned embed w/ ar preserving. Default res to 384x384 to match default seq len.

rwightman

a problem about pre-trained weights.

6

I found a problem about pre-trained weights. When I trained with a custom model and put it into the same model with the same head, when I used --pretrained-path to...

chenyanting1

Support gradient checkpointing in `forward_intermediates()`

1

Summary: - Renamed `mask` to `attn_mask` in NaFlex ViT `forward_intermediates()` - Removed unused imports - Support gradient checkpointing in `forward_intermediates()`", resolve #2435

brianhou0208

Naflex performance

8

Hi! Thank you for the excellent library and the recent addition of NaFlex transformers. I've been learning a lot from exploring it! While experimenting with the code, I identified a...

stas-sl