mlx-swift-lm

Make Gemma 4 per_layer_model_projection quantizable (#309)

Optimize shared MoE combine paths (#324)

Fix MROPE, rope_deltas, and invFreq loading in Qwen2.5-VL (#239)

Fix Gemma4Text quantized KV cache attention (#237)

Fix sporadic bare-JSON tool call handling and harden JSON parser recovery (#205)