Commit History
24 days ago
-
mac_arm64: MLX and Core ML rebuilt from the released weights (model.onnx f6c74601) — both were still the 2026-08-20 weights; Core ML is F16 again
Verified signaturefd8672c
-
coreml_ane: W8A16 embed package (int8 weights, fp16 activations) — passed a 384-query paired gate, 96 -> 73 MB
Verified signatureee119ed
-
compiled artefacts from mac_arm64 (cpu)
Verified signature0bd77df