Birch-san

Nested tensors

6

Hi, thanks for the library; we're using NATTEN currently for image generation and it's really helpful. I was wondering whether you had any plan to (or already do) support [nested...

Improve performance via batched-matmul and fused multiplies

10

Many thanks for providing this reference implementation. I tried integrating this into stable-diffusion / diffusers. A fix was required to make it work on Mac (PyTorch MPS backend): https://github.com/Birch-san/diffusers/pull/1/commits/04372140a25d7f53549175f1f196599c3e9bf3a5 Knowing...

Any prediction whether this will work for "Rise of the Rajas" expansion?

4

[Rise of the Rajas](http://store.steampowered.com/app/488060/) expansion launches in 4 days (i.e. December 19). Am I right to assume that this is likely to be "more of the same", and as such...

Passes allclose() comparison with original attention.

use torch's built-in Multihead attention class for self-attention

3

CompVis stable-diffusion [uses](https://github.com/CompVis/stable-diffusion/blob/69ae4b35e0a0f6ee1af8bb9a5d0016ccb27e36dc/ldm/modules/attention.py#L152) lucidrains' multi-head attention implementation [from `perceiver_pytorch`](https://github.com/lucidrains/perceiver-pytorch/blob/abbb5d5949d3509c57749bd134f5068f2761aac7/perceiver_pytorch/perceiver_pytorch.py#L82). there's actually a built-in [MultiheadAttention](https://pytorch.org/docs/stable/generated/torch.nn.MultiheadAttention.html) class in torch that we could be using. it even has a fast-path for self-attention, which...

No attempt at artefact exfiltration, version number inference, release tagging

Arbitrary masks

12

Congratulations on shipping FNA backward! Looking forward to using it. On another note: would it be possible to support arbitrary masking? MaskDiT outperformed regular DiT, with a 70% reduction in...

_Note: for a cleaner diff: merge-base of this PR is set to an unrelated branch in my repository, rather than upstreaming to [crowsonkb/k-diffusion](https://github.com/crowsonkb/k-diffusion). because this branch is downstream of my...

```bash python -m clip_guided_diffusion.main 'A beautiful fantasy painting of the wind by Wojciech Siudmak' \ --checkpoint /sdb/ckpt/imagenet_test_v2_007_02200000.safetensors \ --config config/hdit_557m.json \ --seed 436376137 \ --size 256 256 ``` ![00003_436376137_ 2000...