leiwen83

W4A4 precision

3

Hi, I tried quant resnet50 in w8a8 mode, and it achieve good result to 77%, but when I switch to test W4A4, using below command, its accary drop to :...

Dyadic number

1

Hi, I check current code base, it seems to me current scale factor is still using linear mapping, which is used in SymmetricQuantFunction? So where Dyadic number is used for...

Would spynet be added?

2

spynet is a tiny optical flow which compute efficient, and it is also used by like model like basicVSR. Would it be added to mmflow?

add Conv singlestage

11

This commit set add singlestage conv for cutlass

AdderNet for SR

1

Hi, I see recent you have applying addernet for EDSR quantization, which is a great work. But for some other SR, like EDVR, using deformable conv for the frame alignment....

operation is not registered

4

Hi, I test mace_cc_benchmark over Xiaomi11 ultra phone which is rooted. It is built with the command: bash tools/cmake/cmake-build-arm64-v8a.sh with android ndk:android-ndk-r21e But it reported follow error when I try...

Some questions about int8 quat

1

Hi, I found this repo is very useful for helping understanding trt int8 functions. However, I don't quitly understand the usage for calibration-data which mentioned in the README that it...

Hi, EDVR is a amazing model, but its compution effort is large. Could we use some tech like group convolution to reduce the FLOPS? As https://github.com/DestinyK/MobileSR shows it use "parallel...

Could EDVR be applied with this method?

1

I am interesting wheter video super resolution like EDVR could be applied using your distillation work? Thx, Lei

Internal error when reording ONNX operators

1

Hi, I try to take some test over some onnx model, but meet the error as "internal error when reording onnx operators" Traceback (most recent call last): File "convert.py", line...