Aleksa Gordić

Here is a simple fix: `conda install cudatoolkit-dev=11.3 gxx=10.3 cuda-nvcc=11.3 -c conda-forge -c nvidia` https://github.com/gordicaleksa/Open-NLLB/blob/nllb_replication/INSTALL.md

yeah hitting this same one - completely unintuitive behavior

Surprised no one mentioned alibi, there are ways to train the model at size X during the training and it can extrapolate to Y > X during inference (alibi being...

I'm hitting this same issue on Ubuntu, 48GB of VRAM (2xRTX3090), using OPT-125M model. I'm comfortably using large ML systems on my rig in general. the pin memory hack works...

Ok, I've done some debugging, this function is literally designed to allocate GPU memory as max as possible (up to what's set in the `gpu_memory_utilization`). What's the reasoning behind this...

Hi @guillaumekln! How hard would it be to implement this? Can you maybe give us some pointers? Most of the popular libraries support v1.0 already

1. Your config is not good, remove those 2 directions and add `eng_Latn-rus_Cyrl` (or more likely I think only `eng-rus` will work). Same for the opposite direction that you need....

Hitting the same issue with `imagenet64_uncond_100M_1500K.pt` and 4k steps. Running with fewer diffusion steps (200, 500, 800, 1000) seems to work. I'll get back to the thread if I find...

Hey @Bing1002 if you're not facing this problem any more feel free to close the issue, try and rerun python script and then run your C script again and it...

Hey @PeterZhizhin feel free to close this issue, the +1 is not a bug because it is used only in target when you load a first batch and in the...