daniel/h3-blackwell-runtime
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix causal temporal padding to match reference (2k-1 front zeros when spatial_padding>0)

Browse source
This commit is contained in:
Daniel Maddern 2026-08-19 21:09:03 +07:00
parent 390135fca6
commit 06fd79a8fd
1 changed files with 14 additions and 15 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

29
src/h3_blackwell_runtime/vae_encoder.py
View file

@ -41,24 +41,23 @@ NIN_LEVELS = frozenset({1, 3, 5})
DOWNSAMPLE_LEVELS = frozenset({0, 1, 2, 3}) DOWNSAMPLE_LEVELS = frozenset({0, 1, 2, 3})
def _causal_front_padding(t_in: int, kernel: int, stride: int) -> int:
"""Front-only zero padding: ``kernel-1-(ceil(t_in/stride)-1)*stride``."""
t_out = math.ceil(t_in / stride)
return max(0, kernel - 1 - (t_out - 1) * stride)
def _causal_conv3d(x, weight, bias, *, kernel_size, stride, spatial_padding): def _causal_conv3d(x, weight, bias, *, kernel_size, stride, spatial_padding):
"""Reflect-spatial / causal-temporal 3D conv (single-frame -> truncated taps).""" """Reflect-spatial / causal-temporal 3D conv (matches upstream_vae.CausalConv3d).
Front-only zero temporal padding of ``(kernel_size - 1)`` (two taps on each
side) whenever ``spatial_padding > 0`` (the reference pads with ``(k//2,
k//2, 0)`` in time), no temporal padding otherwise. A single input frame
has its temporal taps truncated instead of convolving zero frames
(``autopad="causal_zero"``) so a keyframe never leaks zero into the latent.
"""
t = x.shape[2] t = x.shape[2]
if t == 1: if t == 1 and spatial_padding > 0:
if spatial_padding > 0: half = (kernel_size - 1) // 2
half = (kernel_size - 1) // 2 kernel = weight[:, :, half : half + 2 * spatial_padding + 1]
kernel = weight[:, :, half : half + 2 * spatial_padding + 1] return F.conv3d(x, kernel, bias, (1, stride[1], stride[2]), (0, spatial_padding, spatial_padding))
return F.conv3d(x, kernel, bias, (1, stride[1], stride[2]), (0, spatial_padding, spatial_padding)) if spatial_padding > 0:
x = F.pad(x, (0, 0, 0, 0, kernel_size - 1, 0))
return F.conv3d(x, weight, bias, stride, (0, spatial_padding, spatial_padding)) return F.conv3d(x, weight, bias, stride, (0, spatial_padding, spatial_padding))
front = _causal_front_padding(t, kernel_size, stride[0])
if front > 0:
x = F.pad(x, (0, 0, 0, 0, front, 0))
return F.conv3d(x, weight, bias, stride, (0, spatial_padding, spatial_padding)) return F.conv3d(x, weight, bias, stride, (0, spatial_padding, spatial_padding))