5d0077ae93
refactor(cache): bind attention layers to cache groups from the plan
2026-09-07 12:20:50 +00:00
69fa679dd2
refactor(attention): simplify QSA verification lifecycle
d91612ec91
refactor(attention): split QSA into paged leaf and shared runtime
比较 2 提交 »
2026-09-07 12:20:49 +00:00
3bac782445
fix(runtime): keep inline output IDs free of detokenizer context (#1419)
2026-09-07 12:20:49 +00:00
2026-09-07 11:58:50 +00:00
2026-09-07 11:58:50 +00:00
089d9d26d6
test: synchronize cold-start timeout waiters before releasing fetch
09fa491ae9
docs: prepare v3.32.0 release
比较 2 提交 »
2026-09-07 11:58:22 +00:00
2026-09-07 11:58:22 +00:00
2026-09-07 11:56:22 +00:00
2026-09-07 11:56:22 +00:00
2026-09-07 11:56:21 +00:00
2026-09-07 11:55:52 +00:00
2026-09-07 11:55:52 +00:00
83f0752cf1
Merge pull request #1266 from TheCrazyAnt/pr/node-beside-cli
87a01c25f5
fix(mac): prefer the node beside the CLI over a version manager's default
比较 2 提交 »
2026-09-07 11:49:54 +00:00
2026-09-07 11:49:45 +00:00
2026-09-07 11:49:35 +00:00
2026-09-07 11:45:09 +00:00