17 Commits

Author SHA1 Message Date
leejet
6dcb5bbd42
fix: handle GPU memory reports and LLM encoding failures (#2020) 2026-09-22 00:28:52 +08:00
leejet
275ab58e01
perf: reduce CPU overhead in graph execution and sampling (#1997) 2026-09-19 18:32:58 +08:00
leejet
9a977388a8
fix: guard GPU memory capacity and propagate encoding failures (#1958) 2026-09-13 21:35:46 +08:00
leejet
6b47fec013
refactor: unify model source and weight lifecycle management (#1956) 2026-09-10 23:53:48 +08:00
leejet
dbb611264e
feat: add verbose logging and log-level selection (#1941) 2026-09-06 23:14:03 +08:00
leejet
462d675018
refactor: unify runner lifecycles and weight residency (#1940) 2026-09-06 22:30:45 +08:00
assouan
6c57cc3b38
feat: prefetch streamed layers during compute (#1905)
Co-authored-by: leejet <leejet714@gmail.com>
2026-09-06 16:35:30 +08:00
leejet
e92e86fb11
fix: prevent torch checkpoint offset overflow (#1832) 2026-07-29 23:16:29 +08:00
yzyyzyhhh
2993b7fb43
fix: make parameter loading backend-aware (#1828) 2026-07-29 22:18:44 +08:00
fszontagh
12b6fbff28
feat: hot-reload ControlNet - swap without rebuilding the context (#1768) 2026-07-10 23:21:38 +08:00
leejet
9ef6e7398f
feat: drive layer split from graph-cut segments (#1762) 2026-07-07 23:16:52 +08:00
Piotr Wilkin (ilintar)
68f3d6df9f
feat: support for cross-device row split (#1735) 2026-07-04 16:31:18 +08:00
leejet
f440ad9c29
fix: avoid writable mmap for read-only weights (#1698) 2026-06-23 00:39:31 +08:00
fszontagh
787d229d84
perf: --eager-load to pre-load params at model-load time (#1687) 2026-06-22 22:10:09 +08:00
leejet
6e66a1a4a4
fix: allow oversized Vulkan parameter tensors (#1662) 2026-06-15 23:18:52 +08:00
leejet
bdb431ad95
feat: support disk params backend (#1651) 2026-06-14 14:48:50 +08:00
leejet
563137a592
refactor: centralize runner weight staging and cleanup (#1644) 2026-06-13 13:19:13 +08:00