SystemErrorWang

@calmevtime it seems the function "block_composition" is necessary in UpsampleDCT. would you please upload the dct_resize.py file, and also clarify the meaning of L, M, and N?

could anyone find "dct_resize.py"? it was imported everywhere but can't be found

> > > > V100 32G能不能训起来？ > > > > > > > > > deepspeed可以降低stage1到33g左右，两卡。如果用更多的卡可以继续降一点。可以考虑用colossalai来卸载大部分参数和优化器状态到cpu上。但我没有尝试成功。。 > > > > > > 您好。请问您可以提供将stage1降到两卡33G的 deepspeed版的代码吗？非常感谢 > > 可以参考摩尔线程的训练代码，通过accelerate启动deepspeed hello，我用摩尔线程的训练代码，deepspeed训练4步之后会卡住不动。请问您是用他们的代码+deepspeed成功了吗？可以介绍一下是怎么用的吗？

> > you can use deepspeed to reduce gpu memory > > i did, but the process will stuck at backward in 5th step. i was confused same problem，training with...

I succeeded in testing on a single image with the provided pre-trained model, but found severe checkerboard artifacts. It may be because of the image pre-process.

same here, my training got stuck after 4 steps, and showing this info: `[loss_scaler.py:183:update_scale] [deepspeed] OVERFLOW! Rank 0 Skipping step. Attempted loss scale: 536870912, reducing to 268435456 `

> 虽然可以发布我们的 csv，但是我们在处理的过程中，没有记录视频剪切的对应时间戳。这样即使发布了可能大家也没法用，所以暂时先不打算发布了。 同求一个pixels的视屏链接，没有对应时间戳也可以自己下载清洗一下

sorry for late reply, i also found this problem in my training. I suggest you to try deep guided filter as it only use tensorflow default ops and no additional...

same problem here, would be really grateful if anyone has a solution

> @thbupt You can try: Follow the steps: > > conda install -c pytorch pytorch=1.9.1 torchvision cudatoolkit=10.2 conda install -c fvcore -c iopath -c conda-forge fvcore iopath conda install -c...