model card: add the int8 build and state what it costs

per-layer quantization parameter table for the int8 build

add the int8 build of the delivered ColorSort policy (22/100 vs fp16's 26/100)

model card: rewrite around the two acts; drop the delisted generations

delist retired generations: single-task v6 (fp16+int8), superseded v13cs/v14cs, guided-decoding task chain

Upload colorsort/grounder_r4_mix20.pt with huggingface_hub

Upload colorsort/v15cs_frozen_fp16.pt with huggingface_hub

Upload v8_mt_colorsort_fp16.pt with huggingface_hub

Delete rl_expert_teacher.pt with huggingface_hub

Delete v6_dino_w4a8.pt with huggingface_hub

Delete v5_lang_fp16.pt with huggingface_hub

Delete v4_fp16.pt with huggingface_hub

Delete ablation/v6_w4a8_clip.pt with huggingface_hub

Delete ablation/v6_w4a4_rtn.pt with huggingface_hub

Delete ablation/v6_w4a4_dyn.pt with huggingface_hub

Delete ablation/v6_w4a4_clip.pt with huggingface_hub

Delete ablation/v6_sub800_fp16.pt with huggingface_hub

Delete ablation/v6_sub3200_fp16.pt with huggingface_hub

Delete ablation/v6_sub200_fp16.pt with huggingface_hub

Delete ablation/v6_int8_static_pc.pt with huggingface_hub

Delete ablation/v5_int8_weightonly.pt with huggingface_hub

Delete ablation/v5_int8_static_pg64.pt with huggingface_hub

Delete ablation/v5_int8_static_pc.pt with huggingface_hub

Delete ablation/v4_int8_static_pg64.pt with huggingface_hub

Delete ablation/v4_int8_static_pc.pt with huggingface_hub

Delete ablation/v3_fp16.pt with huggingface_hub

Delete ablation/v2_fp16.pt with huggingface_hub

Delete ablation/v1_fp16.pt with huggingface_hub

Upload v7_mt_generalist_fp16.pt with huggingface_hub

Upload colorsort/v14cs_fp16.pt with huggingface_hub

Upload colorsort/v13cs_fp16.pt with huggingface_hub

Upload folder using huggingface_hub

initial commit