SceneWorks/scail2-mlx

Add q4/q8/bf16 quant-matrix tier subdirs (sc-9944, epic 8506)

sc-5445: ship lean pre-quantized Q4 snapshot (Q4 DiT 32.8->8.9GB + quantization manifest; prune redundant raw pickles; refresh card)

sc-5445: converted umt5-xxl encoder (bf16) for Umt5Encoder

sc-5445: converted open-CLIP ViT-H/14 visual tower (f32) for ScailClip

sc-5445: converted Wan2.1 z16 VAE (f32) for WanVae

sc-5445: root umt5 tokenizer.json for Umt5Encoder

sc-5445: turnkey MLX model card (converted vae/t5/clip safetensors)

Add umt5-xxl/special_tokens_map.json

Add umt5-xxl/tokenizer_config.json

Add umt5-xxl/tokenizer.json

Add umt5-xxl/spiece.model

Add umt5-xxl/models_t5_umt5-xxl-enc-bf16.pth

Add bias-aware-dpo-lora.pt

Add models_clip_open-clip-xlm-roberta-large-vit-huge-14-onlyvisual.pth

Add Wan2.1_VAE.pth

Add dit.safetensors

Add config

Add model card

initial commit