xinyun/ComfyUI
1
0
Fork 0
mirror of https://git.datalinker.icu/comfyanonymous/ComfyUI synced 2026-09-13 07:27:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix mask shapes

Browse Source
This commit is contained in:
Raphael Walker 2024-12-16 10:22:12 +01:00
parent 5e8206362c
commit b321bfc8d3
1 changed files with 4 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
comfy/ldm/modules/attention.py
View File

@ -378,17 +378,16 @@ def attention_xformers(q, k, v, heads, mask=None, attn_precision=None, skip_resh
# add a singleton batch dimension # add a singleton batch dimension
if mask.ndim == 2: if mask.ndim == 2:
mask = mask.unsqueeze(0) mask = mask.unsqueeze(0)
if mask.ndim != 3:
raise ValueError(f"Bad mask shape {list(mask.shape)}. Valid shapes are [b, Nq, Nk], [1, Nq, Nk], or [Nq, Nk].")
# add a singleton heads dimension # add a singleton heads dimension
if mask.ndim == 3:
mask = mask.unsqueeze(1) mask = mask.unsqueeze(1)
# pad to a multiple of 8 # pad to a multiple of 8
pad = 8 - mask.shape[-1] % 8 pad = 8 - mask.shape[-1] % 8
# the xformers docs says that it's allowed to have a mask of shape (1, Nq, Nk) # the xformers docs says that it's allowed to have a mask of shape (1, Nq, Nk)
# but when using separated heads, the shape has to be (B, H, Nq, Nk) # but when using separated heads, the shape has to be (B, H, Nq, Nk)
# in flux, this matrix ends up being over 1GB # in flux, this matrix ends up being over 1GB
# Therefore we create a mask with a singleton heads dimension, and then expand it (ie, create a view) # here, we create a mask with the same batch/head size as the input mask (potentially singleton or full)
mask_out = torch.empty([mask.shape[0], 1, q.shape[1], mask.shape[-1] + pad], dtype=q.dtype, device=q.device) mask_out = torch.empty([mask.shape[0], mask.shape[1], q.shape[1], mask.shape[-1] + pad], dtype=q.dtype, device=q.device)
mask_out[..., :mask.shape[-1]] = mask mask_out[..., :mask.shape[-1]] = mask
# doesn't this remove the padding again?? # doesn't this remove the padding again??