YJYJLee

Is it possible to train Cylinder3D with mixed precision?

2

Hello, I am trying to train Cylinder3D with mixed precision, so I added `torch.cuda.amp` code to the source code. However, I am getting NaN value due to overflow as soon...

Installing MinkowskiEngine as develop mode (pip install --editable .)

9

Hi, I am trying to build [email protected] as develop mode for the research purpose. I tried the command `python setup.py install` and succeed in installing it and checked MinkowskiEngine being...

Can you upload the processor code(run_classifier.py) for glue dataset(cola, qqp, sst-2, rte, mrpc)?

1

Hi, I would like to run xlnet on glue dataset, but I found out that run_classifier.py has only processor code for mnli and sts-b. Can you upload your processor codes...

How to support key_padding_mask

1

# ❓ Questions and Help Hi, I am trying to replace existing `torch.nn.functional.scaled_dot_product_attention` with `xformers.ops.memory_efficient_attention`. I want to use `xformers.ops.fmha.flash` and `xformers.ops.fmha.triton` as ops, but I am struggling with how...

Missing key(s) and size mismatch error for SeamlessM4T_medium

8

Hi, I am trying to use SeamlessM4T_medium ckpt for evaluation, but I am getting following error while loading the ckpt. I just added `--model_name seamlessM4T_medium` to the command, is there...

PR request for Pytorch blog post. Summary: This post is the fourth part of a multi-series blog focused on how to accelerate generative AI models with pure, native PyTorch. In...

Most effective way of applying key_padding_mask

1

Hi, I am trying to integrate flash-attention into the model I am looking at. My model uses `key_padding_mask` to support variable size of samples in a batch during finetuning. I...

PR request for Pytorch blog post. Summary: This post is the fourth part of a multi-series blog focused on how to accelerate generative AI models with pure, native PyTorch. In...