refactor: share the HF cache helpers between qwen and sglomni2026-09-04T21:41:14+00:00historiahistoriavg@proton.me2026-09-04T21:41:14+00:005263a30356d7a7b39490e9a3cf5f6c179249500c
feat: pip install checks for bad crc errors and purges cache automatically2026-09-03T21:26:14+00:00historiahistoriavg@proton.me2026-09-03T21:26:14+00:009dd66997033f7d306718c67745b41c16dd034865
fix: increase gpu memory for MOSS with gplang-omni to .872026-09-03T05:36:45+00:00historiahistoriavg@proton.me2026-09-03T05:36:45+00:00976f12a72ebbd330e0328991afed3524c06a35ad
feat: gplang-omni reserves 70% of gpu 0 memory, preventing oom errors on consumer cards2026-09-03T04:38:47+00:00historiahistoriavg@proton.me2026-09-03T04:38:47+00:007ec70e19b78cbe7438ec4d61e15951f6e8dd0fb8
fix: warn before using a likely too-big chunk size for sglang-omni models2026-09-03T02:53:07+00:00historiahistoriavg@proton.me2026-09-03T02:53:07+00:00d04a2c53b926ccde0d582dbf4a7360dc0f072205
feat: reserve 20% vram when running higgs with sglang-omni2026-09-03T00:46:42+00:00historiahistoriavg@proton.me2026-09-03T00:46:42+00:007a7dca313750ee75e0f8a2a5442ca5d78e743294
feat: sglang-omni model picker on manual server launch2026-09-02T23:13:24+00:00historiahistoriavg@proton.me2026-09-02T23:13:24+00:006804c785c6b506c47b45264398728d0a609310be
fix: ZONOS2 only generating 1024 tokens max2026-09-02T22:20:14+00:00historiahistoriavg@proton.me2026-09-02T22:20:14+00:00f14b80f6b42a0d0891718eb6284336b1694795cf
feat: auto-update via settings2026-09-02T21:16:05+00:00historiahistoriavg@proton.me2026-09-02T21:16:05+00:00717a3dc112519ae7cf39a212a4e3ec9e1ba17f28