Patrick Devine 82e0ddb6fe mlxrunner: harden linear/embedding layers against over-promotion (#16682) ...

Adding/Multiplying a tensor by a scalar w/ a different data type
can cause the tensor to be promoted and cause performance issues.

This change adds several guards against over-promotion.

2026-06-11 13:56:25 -07:00

..

gemma3

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

gemma4

mlxrunner: drive MTP speculation through cache snapshots

2026-06-09 00:39:19 -07:00

glm4_moe_lite

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

laguna

mlxrunner: move YaRN RoPE helpers into x/models/nn

2026-05-22 09:32:09 -07:00

llama

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

nn

mlxrunner: harden linear/embedding layers against over-promotion (#16682)

2026-06-11 13:56:25 -07:00

qwen3

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

qwen3_5

mlxrunner: drive MTP speculation through cache snapshots

2026-06-09 00:39:19 -07:00

qwen3_5_moe

MLX: add header vendoring and remove go build tag (#14642)

2026-03-09 17:24:45 -07:00