Warren Lo

warren pushed to v2 at warren/markbaseengine

2026-07-06 02:53:52 +08:00

96fe213bc4 v2: add E4B multimodal test, fix VisionTower missing groupSize

warren pushed to v2 at warren/markbaseengine

2026-07-06 01:31:35 +08:00

97f9bdcf90 v2: add full context 2048-token, repeated tokens, edge token tests

warren pushed to v2 at warren/markbaseengine

2026-07-06 01:11:51 +08:00

16c16b9bee v2: add 1024-token long context test

warren pushed to v2 at warren/markbaseengine

2026-07-06 01:01:45 +08:00

7e686c3c5a v2: add long context 12B test (256 tokens)

warren pushed to v2 at warren/markbaseengine

2026-07-05 23:58:45 +08:00

af1d10737e v2: add multimodal 12B test, fix VisionTower12B kernel dispatch

warren pushed to main at warren/momentry_core

2026-07-05 23:36:59 +08:00

cb604b74ec docs: update Rule 1 OCR independent chunks documentation

warren pushed to v2 at warren/markbaseengine

2026-07-05 23:33:54 +08:00

07459e8ee3 v2: add 12B model test (Model12BTest)

warren pushed to main at warren/momentry_core

2026-07-05 23:31:09 +08:00

e91d51cc5e feat: OCR independent chunks (方案 A) + stats API

warren pushed to main at warren/momentry_core

2026-07-05 23:28:11 +08:00

b3407c83c7 feat: OCR independent chunks (方案 A)

warren pushed to v2 at warren/markbaseengine

2026-07-05 22:41:50 +08:00

7a8edf77ee v2: remove remaining logit scaling hacks from batch/optimized paths

warren pushed to main at warren/momentry_core

2026-07-05 22:22:24 +08:00

5a3f791ecd feat: only run identity agent if file has seed identities

warren pushed to main at warren/momentry_core

2026-07-05 22:06:38 +08:00

0b82aa875c feat: Rule 1 now creates chunks for OCR-only text

warren pushed to v2 at warren/markbaseengine

2026-07-05 19:52:57 +08:00

239474bef0 v2: fix 26B activation explosion — normalize groupSize=32 scales, fix hardcoded loops

warren pushed to main at warren/momentry_core

2026-07-05 19:49:07 +08:00

465552f8b2 fix: remove duplicate 'asr' pre_chunks storage in ASRX handler

warren pushed to v2 at warren/markbaseengine

2026-07-05 16:12:16 +08:00

8a29dae613 v2: add 26B + 31B model tests (Phase 3)

warren pushed to v2 at warren/markbaseengine

2026-07-05 15:05:10 +08:00

2fd03d0ac1 v2: fix GPU non-determinism test tolerance

warren pushed to v2 at warren/markbaseengine

2026-07-05 14:52:20 +08:00

e9ab994533 v2: add E4B-MarkBase model tests (Phase 2)

warren pushed to v2 at warren/markbaseengine

2026-07-05 13:58:25 +08:00

97798850e3 v2: clean up CI test triggers

46b2e5382b ci: trigger v1.0.8 runner test

5d1b2df0f1 ci: trigger test run

Compare 3 commits »

warren pushed to v2 at warren/markbaseengine

2026-07-05 13:41:51 +08:00

31427770b1 v2: Apply tokenizer UTF-8 fix + Engine writeFloats helper

5a94501f95 Add bf16 layer weight support for E4B model

Compare 2 commits »

warren pushed to v2 at warren/markbaseengine

2026-07-05 13:31:47 +08:00

e23ef405bc v2: Fix build conflict and unit tests