Oct 14 10:30 Model card: give the table real headers Oct 13 20:22 Shape folders: metadata, slot layout, tokenizer, autoencoders per shape (imagine 0.12) 20:20 Shape folders: metadata, slot layout, tokenizer, autoencoders per shape (imagine 0.12) 20:17 Shape folders: metadata, slot layout, tokenizer, autoencoders per shape (imagine 0.12) 20:15 Shape folders: metadata, slot layout, tokenizer, autoencoders per shape (imagine 0.12) 20:14 4:3 edit transformer with the LoRA slot 20:12 4:3 t2i transformer with the LoRA slot 20:05 2x3 edit transformer with the LoRA slot 20:03 3x2 edit transformer with the LoRA slot 20:01 3x4 edit transformer with the LoRA slot 19:59 9x16 edit transformer with the LoRA slot 19:57 16x9 edit transformer with the LoRA slot 19:56 3x2 t2i transformer with the LoRA slot 19:54 4x5 t2i transformer with the LoRA slot 19:53 9x16 t2i transformer with the LoRA slot 19:51 16:9 text-to-image transformer with the LoRA slot (bit-identical to aspect-16x9's without a person) 19:46 1:1 editing transformer with the LoRA slot (bit-identical to the plain one without a person) 16:24 Name the LoRA-slot transformer TransformerSlot.aimodel (imagine 0.11.0): a binary from before the slot finds no Transformer.aimodel and fetches the plain one from its own pinned commit instead of failing on the slot's inputs 14:38 Transformer.aimodel is now the LoRA-slot export (rank 16): zero LoRA inputs = the plain transformer bit for bit; slot.json names the layers One transformer for plain runs and personal LoRAs (imagine --person). Exported with arraypress/coreai-models 1.0.0-nonsquare.5 `--lora-slot 16`. Earlier revisions keep the plain export. 14:37 lora-slot-r16: FLUX.2 Klein 4B transformer with LoRA matrices as inputs (rank 16) Oct 12 17:06 Model card: 9:16 and 4:5 shapes 17:05 Add 4:5 (896 × 1120): transformer + decoder for this shape; shares ../TextEncoder.aimodel 17:04 Add 9:16 (768 × 1344): transformer + decoder for this shape; shares ../TextEncoder.aimodel Oct 11 19:04 16:9 (1344 x 768) export in aspect-16x9/, and the card says how shapes cost 17:52 Core AI conversion, model card, licences and NOTICE 17:52 initial commit