DarthAffe/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-09-24 20:20:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: enable VAE decode tiling fallback without auto-fit (#1932)

Browse Source
This commit is contained in:
Hmission 2026-09-11 01:34:56 +08:00 committed by GitHub
parent 14eddb32b1
commit b68d58624d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
docs/backend.md
View File

@ -188,8 +188,9 @@ weights, compute buffers and caches must
still fit the runner's capacity checks. Offloading weights does not guarantee still fit the runner's capacity checks. Offloading weights does not guarantee
that every resolution or frame count will fit, and auto-fit does not change a that every resolution or frame count will fit, and auto-fit does not change a
component to CPU computation solely because its full weights exceed VRAM. component to CPU computation solely because its full weights exceed VRAM.
If a VAE decode fails, auto-fit retries with spatial tiling; supported video If a VAE decode fails, decoding retries with spatial tiling even when `--auto-fit`
decoders try temporal tiling first and can then add spatial tiling. is off; supported video decoders try temporal tiling first and can then add
spatial tiling. Spatial retries use half-size tiles along each latent dimension.
## Modules ## Modules

4
src/core/backend_fit.cpp
View File

@ -390,6 +390,8 @@ namespace sd::backend_fit {
retry_mode = tiling_params.enabled ? "spatial+temporal" : "temporal"; retry_mode = tiling_params.enabled ? "spatial+temporal" : "temporal";
} else if (!tiling_params.enabled) { } else if (!tiling_params.enabled) {
tiling_params.enabled = true; tiling_params.enabled = true;
tiling_params.rel_size_x = 0.5f;
tiling_params.rel_size_y = 0.5f;
if (tiling_params.tile_size_x <= 0) { if (tiling_params.tile_size_x <= 0) {
tiling_params.tile_size_x = 256; tiling_params.tile_size_x = 256;
} }
@ -401,7 +403,7 @@ namespace sd::backend_fit {
return false; return false;
} }
LOG_WARN("auto-fit: VAE decode failed (likely out of memory); retrying with %s tiling", LOG_WARN("VAE decode failed (likely out of memory); retrying with %s tiling",
retry_mode); retry_mode);
return true; return true;
} }

1
src/pipeline/diffusion_engine.cpp
View File

@ -2621,7 +2621,6 @@ sd::Tensor<float> StableDiffusionGGML::decode_first_stage(const sd::Tensor<float
auto decoded = first_stage_model->decode(n_threads, latents, vae_tiling_params, decode_video, circular_x, circular_y); auto decoded = first_stage_model->decode(n_threads, latents, vae_tiling_params, decode_video, circular_x, circular_y);
const bool prefer_temporal_tiling = decode_video && first_stage_model->can_temporal_tile_decode(); const bool prefer_temporal_tiling = decode_video && first_stage_model->can_temporal_tile_decode();
while (decoded.empty() && while (decoded.empty() &&
auto_fit_enabled &&
sd::backend_fit::prepare_vae_decode_retry_tiling(vae_tiling_params, prefer_temporal_tiling)) { sd::backend_fit::prepare_vae_decode_retry_tiling(vae_tiling_params, prefer_temporal_tiling)) {
decoded = first_stage_model->decode(n_threads, latents, vae_tiling_params, decode_video, circular_x, circular_y); decoded = first_stage_model->decode(n_threads, latents, vae_tiling_params, decode_video, circular_x, circular_y);
} }