mlboydaisuke/Gemma-4-12B-CoreAI
Card: macOS/iOS 27 GA wording (beta requirement dropped; beta findings dated)
gemma4_12b_qat_decode_int4linsym_msdpa_g8: re-save with coreai-core 1.0.0b2 (strip debug locations; weights unchanged). The 0.4.0-era IR refused AIModel.load on every OS 27 build since beta 2, still on the RC 26A428 (2026-09-14); the re-saved bundle loads. Recovery: coreai-model-zoo conversion/recovery/{strip_b1,resave_b2}.py
gemma4_12b_qat_decode_int8lin_msdpa_g8: re-save with coreai-core 1.0.0b2 (strip debug locations; weights unchanged). The 0.4.0-era IR refused AIModel.load on every OS 27 build since beta 2, still on the RC 26A428 (2026-09-14); the re-saved bundle loads. Recovery: coreai-model-zoo conversion/recovery/{strip_b1,resave_b2}.py
gen-cards: regenerate Use-it block
Card: DeviceMark row (2026-09)
Card: definition line (Core AI, 2026-09)
Match the License section to the base card (Apache License 2.0)
Card: base_model_relation: quantized (so the conversion lists under the base model's Quantizations, not Finetunes)
Card: license metadata now matches the base model card (apache-2.0 + Gemma 4 license link)
gen-cards: regenerate Use-it block
Link the card back to its collection and the request box
Add config.json so the Hub can count downloads
Take Google's 2026-07-09 canonical chat template
gen-cards: regenerate Use-it block
gen-cards: regenerate Use-it block
gen-cards: add Use-it markers
chat-EOS fix: eos_token -> <turn|> (clean stop in generic apps)
chat-EOS fix: eos_token -> <turn|> (clean stop in generic apps)
Gemma 4 12B: drop simple-kernel gemma4_12b_qat_decode_int8lin_msdpa (superseded by _g8)
Gemma 4 12B: gemma4_12b_qat_decode_int4linsym_msdpa_g8 (higher-occupancy decode bundle)
Gemma 4 12B: gemma4_12b_qat_decode_int8lin_msdpa_g8 (higher-occupancy decode bundle)
Gemma 4 12B: card -> higher-occupancy (_g8) kernel as the ship
Gemma 4 12B: gemma4_12b_qat_decode_int8lin_msdpa (metal-sdpa decode bundle)
Gemma 4 12B: Gemma Terms of Use
Gemma 4 12B: model card
initial commit
