Cream icon indicating copy to clipboard operation
Cream copied to clipboard

This is a collection of our NAS and Vision Transformer work.

Results 44 Cream issues
Sort by recently updated
recently updated
newest added

Thank you for your great job! I have some questions about the shortcut design of FFN in AutoFormer. https://github.com/microsoft/Cream/blob/83a154beb6f85dd6141853b4b7c0738eeec628ba/AutoFormer/model/supernet_transformer.py#L245-L246 To match the dimension setting of FFN, `sample_embed_dim` should be the...

AutoFormer

Hi, first thanks for the nice work! It's really interesting, Currently, I did like to calculate the flops of AutoFormer subnet, however, when I use `flops_count` in fvcore.nn that you...

AutoFormer

Hi, I have searched a submodel (Base) with your default training setting, but it has poor performance(78.3%). How to retrain it and what are the hyper-parameters of retraining setting? Thanks.

AutoFormer

Thank you for sharing your codes. While I try to use your codes, I found something weird. There are performance graphs on README.md of AutoFormerV2. And at the first graph,...

AutoFormerV2

thanks for release the config of autoformerv2! But i have some question about the config. The searched config is so regular. It looks the real searching param is DEPTHS stage3(S3-B...

Thanks for the great work! I am trying to test the evaluation result on the subnet of AutoFormerV2 using the command provided in the repo, however, I cannot find the...

Hi, I haven't found specific architecture details of S3 model in your [**paper**](https://arxiv.org/abs/2111.14725). Similarly, I haven't found model config parameters in github folder Cream/AutoFormerV2. Could you provide specific model paramter...

AutoFormerV2

Hi~ 谢谢你们完美的工作 你能给我提供你们训练超网的LOG文件吗 我发现在训练超网时prec1特别低 如下 ![image](https://user-images.githubusercontent.com/68720156/159231668-e7080110-eb9b-4e01-a316-e3f22435b83e.png) 我在训练spos的时候前6个epoch精度已经到30左右了

Dear Authors, Very impressive work. But when I use the code to save teacher logits, this RuntimeError happened. Any ideas? Traceback (most recent call last): File "main.py", line 595, in...

TinyViT