Nov 15 2 days ago 10:49 Card: shorter top, version history at the end, stale 1.x lines removed 05:38 Model card for radiance 2.0 00:32 Model card for radiance 2.0 Nov 14 3 days ago 23:11 Add tag radiance-2: tested with radiance 2.0.0 Nov 4 13 days ago 16:01 Version 2.0: a4 format (u4 with zero point, f16 scale per 128, int8 activations per 128), calibrated with keptn/quant-calibration-v1. Version 1.0 stays at revision v1.0 Nov 1 16 days ago 02:21 Update model card Oct 24 24 days ago 10:48 Card: BetterBench tables show every category of the run (chat, file_edit, math, reasoning added from the same runs) 09:56 Card: --vision-tower-placement ram needs radiance 1.7.0 or newer Oct 19 29 days ago 16:07 Model card: shorter sentences, standard benchmark intros, no content change Oct 14 06:18 Model card: vision tower in RAM with PR #91, device numbers 05:22 Model card: vision tower in RAM is not usable yet 05:04 Model card: highlights, vision, memory next to the vision tower, speed on 1/2/4 cards Oct 13 23:31 Model card: vision benchmarks (lmms-eval) 21:31 Add files using upload-large-folder tool 21:14 v1.0: SmoothQuant on q/k/v, about a third closer to the original; radiance 1.5.0 Oct 12 00:34 Add config.json as download counter Oct 11 23:01 Card: requirements are #75 and #81 22:52 Card: shorter experimental notice 22:51 Card: experimental notice 22:50 Card: requirements without #83 (included in #81) 15:10 gemma4-31b-w4a8: radiance container