EfficientFormer issues

Accuracy from distillation?

1

Thanks for the great work. One question regarding accuracy: are the numbers from distillation or not? The release model has distillation head so I wonder what is the number without...

fyangf

Models in TorchHub

1

Is there any plan on adding the pretrained weights in TorchHub?

nahidalam

Error in conversion from ONNX to TFLite

I used the EfficientFormer as backbone for training a face embedding model and the accuracy was pretty good. But I tried to convert it to tflite and deploy on edge...

TrNgTinh

RuntimeError: Unknown model (efficientformer_l1)

3

I run the code on my computer without GPU, but the error happens at creating model: RuntimeError: Unknown model (efficientformer_l1). The versions of torch, timm, and submitit are 1.8.0+cpu, 0.6.7...

Jialeen

can't convert to coreml

1

when i try to convert to coreml i get ValueError: node input.19 (gelu) got 2 input(s), expected [1] im using pytorch version 1.12.0 and coremltools version 5.2.0

rotem154154

Input size must be 224x224 ?

1

Hello, can we input images of different sizes to the model? When I change the input size as 1024x1024, it incurs some errors.

datong-new

How to apply MetaBlock3D to detection framework?

HI, if I look it right, the 4 stages output shape would be something like this: ``` stage3 torch.Size([1, 448, 16, 16]) stage4 torch.Size([1, 448, 16, 16]) stage5 torch.Size([256, 1,...

lucasjinreal

Are there any the experiments comparing LN and BN?

1

In section 3, Observation 3 part. " Based on our ablation study in Appendix Tab.3, CONV-BN only slightly downgrades performance compared to **LN**" . However I found no **LN** related...

hailong23-jin