Hugging Face's logo

mlboydaisuke
/
gemma-4-E2B-CoreAI

Text Generation
coreai
coreai-aimodel
aimodel
apple-silicon
ane
on-device
gemma-4
custom-metal-kernels
gpu-pipelined
Model card Files
xet
 Community
gemma-4-E2B-CoreAI / macos

Commit History

  1. gemma4_e2b_metal_int8v3_L35.aimodel: 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

    mlboydaisuke
    Verified signature
    f09b3d4

  2. gemma4_e2b_head_argmax_kernel.aimodel: 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

    mlboydaisuke
    Verified signature
    2f9c8a7

  3. gemma4_e2b_frontend_int8.aimodel: 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

    mlboydaisuke
    Verified signature
    60ca97f

  1. macOS best: head+argmax kernel (int8)

    mlboydaisuke
    Verified signature
    4ba4544

  2. macOS best: int8 fused-kernel core (56.6-59 tok/s)

    mlboydaisuke
    Verified signature
    84433ff

  3. macOS best: gather front-end (.aimodel, int8)

    mlboydaisuke
    Verified signature
    517f747