Oct 21 14:09 Add tested LFM2.5 2.6B Vulkan export and extend mobile context to 8K 12:31 Add instruction-tuned LFM2.5 2.6B XNNPACK and MLX exports Aug 27 11:21 Add LFM2.5-VL 1.6B MLX int8 (8w) build Weight-width ablation of the int4 MLX build for software-mansion/react-native-executorch#1379. Same export in every other respect: bf16 compute dtype, group size 32, vision tower unquantized. Aug 16 15:55 Add LFM2.5-VL 1.6B vulkan (8da4w) build 15:39 Add LFM2.5-VL 450M vulkan (8da4w) build Jun 18 12:36 Rename vl_1_6b/mlx/lfm2_5_vl_1_6b_fullmlx_allbf16_dyn.pte to vl_1_6b/mlx/lfm_2_5_vl_1_6b_mlx_int4.pte 12:14 Upload lfm2_5_vl_1_6b_fullmlx_allbf16_dyn.pte 11:57 Rename vl_450m/mlx/lfm2_5_vl_450m_fullmlx_allbf16_dyn.pte to vl_450m/mlx/lfm_2_5_vl_450m_mlx_int4.pte 11:57 Upload lfm2_5_vl_450m_fullmlx_allbf16_dyn.pte Jun 17 12:57 Rename 1_2b/mlx/lfm2_5_1_2b_mlx_4w_v2.pte to 1_2b/mlx/lfm_2_5_1_2b_mlx_int4.pte 12:53 Upload lfm2_5_1_2b_mlx_4w_v2.pte 12:50 Rename 350m/mlx/lfm2_5_350m_mlx_4w_v2.pte to 350m/mlx/lfm_2_5_350m_mlx_int4.pte 12:50 Upload lfm2_5_350m_mlx_4w_v2.pte May 13 14:45 Fix lfm-2.5 layout: distinguish VL from text + colocate tokenizers 13:11 Restructure to MODEL_SPEC.md convention Apr 9 07:28 Upload lfm2_5_vl_450m_8da4w_xnnpack.pte Apr 1 10:16 Upload folder using huggingface_hub 10:15 Upload folder using huggingface_hub 10:15 Upload folder using huggingface_hub 10:15 initial commit