Chat template: accept the 0.18 content-parts form (string form unchanged); weights, tokenizer and executor metadata byte-identical

Card: base_model_relation: quantized (so the conversion lists under the base model's Quantizations, not Finetunes)

Link the card back to its collection and the request box

Add desktop LiteRT-LM CLI section (import/run/serve)

Card: add 'Reproduce (official tools only)' section

Add the blockwise-128 int4 recipe (reproduce with stock litert-torch)

Rebuild via stock litert-torch (recipe.json, no custom code); same blockwise-128 int4, GSM8K 77%

Card: blockwise-128 + cache2048 specs (iPhone ~27 tok/s, GSM8K 77%)

Switch to blockwise-128 int4 + cache2048 (iPhone 27 tok/s, GSM8K 77% parity)

Update card: blockwise int4 + GSM8K parity table

Replace channelwise int4 with blockwise int4 + cache4096 (GSM8K 79% ~ bf16 75%)

Upload model.litertlm with huggingface_hub

Upload README.md with huggingface_hub

initial commit