psychedelicious 91f91aa835 feat(mm): prepare kontext latents before loading transformer ...

If the transformer fills up VRAM, then when we VAE encode kontext
latents, we'll need to first offload the transformer (partially, if
partial loading is enabled).

No need to do this - we can encode kontext latents before loading the
transformer to reduce model thrashing.

2025-08-12 10:51:05 +10:00

..

api

chore: ruff

2025-07-31 06:38:48 +10:00

assets/images

tweaks in response to psychedelicious review of PR

2023-07-26 15:27:04 +10:00

invocations

feat(mm): prepare kontext latents before loading transformer

2025-08-12 10:51:05 +10:00

services

feat(app): add setting to disable picklescan

2025-08-05 09:54:00 +10:00

shared

tidy(nodes): move all field things to fields.py

2024-03-01 10:42:33 +11:00

util

Fix division by zero in multi-diffusion pipeline with creativity values

2025-07-01 15:00:02 +10:00

__init__.py

fix: make invocation_context.py accessible to mkdocs

2024-03-01 10:42:33 +11:00

api_app.py

feat(app): add setting to disable picklescan

2025-08-05 09:54:00 +10:00

run_app.py

Ruff Fix

2025-05-30 19:03:43 +10:00