pytorch-image-models

pytorch-image-models copied to clipboard

Add noisystudent train method

1

How about adding noisystudent training method?

dongfangzhizhu

Fix MultiEpochsDataLoader when not all is consumed

4

If for some reason `MultiEpochsDataLoader.__iter__` gets called but in a previous call the caller didn't consume all its items, then it's gonna continue the previous one, which is incorrect. I...

bryant1410

bryant1410

user3984

rwightman

[BUG] CPU Memory Leakage on TPU

3

**Describe the bug** I find strange memory behaviour with `bits_and_tpu` timm code on tpu. If the layer decay learning rate is set (`--layer-decay`), the memory usage just keep going up....

zw615

Using timm gcvit models and getting errors when using resolution other than 224x224

9

Is there a way to use this with arbitrary resolution? ``` model = timm.create_model('gcvit_tiny', pretrained=True) sample = model(torch.randn(2, 3, 640, 640)) ``` Error ``` --------------------------------------------------------------------------- RuntimeError Traceback (most recent call...

sarmientoj24

Properly get feature maps from Swin-V2 instead of norm

10

Swin-V2 as a backbone gives the feature maps without norm. How to get it?

sarmientoj24

Swin-V2 as backbone for One-Stage Detector has low precision and recall for area==medium

1

I was using Swin-Transformerv2 for object detection as a backbone to a one stage detector. While the metrics are competitive, I am getting very low scores on [email protected]:0.95 because the...

sarmientoj24

[FEATURE] Add Next-ViT

3

Hi, we are a group of engineers from Bytedance Inc. This year, our team published the work: "Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios"(https://arxiv.org/abs/2207.05501) (https://github.com/bytedance/Next-ViT)....

XiaXin-Aloys