 xmarre
|
c121278c65
|
Thread adaptive trim to the target CUDA device
|
2026-04-15 03:05:32 +02:00 |
|
|
xmarre
|
da3f105661
|
Move VRAM trimming into resident loaders
|
2026-04-15 02:57:40 +02:00 |
|
|
xmarre
|
19e69a8ecc
|
Address PR review comments
|
2026-04-15 02:29:52 +02:00 |
|
|
xmarre
|
f7b8fe5cc9
|
Add selective checkpoint ingest and resident VRAM trim
|
2026-04-15 02:17:16 +02:00 |
|
|
xmarre
|
df9128e9ee
|
Fix resident loader reuse edge cases
|
2026-04-13 06:16:06 +02:00 |
|
|
xmarre
|
9f4b27eb23
|
Narrow resident loader hot paths
|
2026-04-13 06:00:19 +02:00 |
|
|
xmarre
|
1b80a6a82b
|
Preserve sparse extra UNet overrides
|
2026-04-12 10:36:40 +02:00 |
|
|
xmarre
|
6d4bd6c7c4
|
Filter extra diffusion weights to UNet keys
|
2026-04-12 10:24:59 +02:00 |
|
|
xmarre
|
24ab29880a
|
Scope loader policy override to load
|
2026-04-12 06:45:32 +02:00 |
|
|
xmarre
|
791a48afd3
|
Add explicit residency policy loader input
|
2026-04-12 06:35:30 +02:00 |
|
|
xmarre
|
a2bff64207
|
Fix loader state restoration and residency bookkeeping
|
2026-04-12 04:35:13 +02:00 |
|
|
xmarre
|
669e6014fd
|
Fix package imports and patch guards
|
2026-04-12 04:14:07 +02:00 |
|
|
xmarre
|
d0fe3f5058
|
Import GPU resident loader implementation
|
2026-04-12 03:46:23 +02:00 |
|