Chen (Cherise) Chen

Not sure if my answer is 100 per cent correct or not. But in WIKI it says: 'Fast b-spline interpolation on a uniform sample domain can be done by iterative...

Can you check your input dim? thanks.

Hi. I found the solution, which took me several hours as I need to learn the CMake command. So to save others' time and have more time enjoy their life,...

I can not reproduce the results on mnist either, using the above config. I got: Loading SVHN dataset. * Prec@1 21.045 mnist_m Loading MNIST_M dataset. * Prec@1 56.951 syn Loading...

> Thanks for the prompt reply. I am using the same version of metaNN (0.1.5) to support second-order gradients computation while using more advanced PyTorch (1.9.1). I thought using a...

Thanks. It now works. The results are below, a bit higher than reported (maybe due to the upgraded pytorch): => loading checkpoint './runs/mnist/Digits/ckpt_mnist.pth.tar' => loaded checkpoint './runs/mnist/Digits/ckpt_mnist.pth.tar' (iter 10001) svhn...

Sorry, I still found it didn't work. The previous post was reported by mistake as they are results from the downloaded checkpoints. If I trained from scratch, I still get...

I did preprocess the data but cannot produce your results. Thanks for sharing us with the processed one. Now, I got much higher performance: Loading SVHN dataset. * Prec@1 35.928...

> I am using PyTorch (1.9.1), MetaNN 0.1.5.

Sure. Can you please kindly share your environment config (requirements.txt) for me to replicate the environment? e.g. you can use pipreqs to generate the dependency list. Thanks.