ollama/mlxrunner/model
Daniel Hiltgen cecac8bf1b
mlxrunner: add Inkling 2 Mini (#19799)
* mlxrunner: add Inkling 2 Mini

* mlx: fuse Inkling 2 Mini residual mixing on Metal

Fuse RMS normalization and query scoring in one kernel. Apply the
softmax weights in a second kernel to avoid storing per-query products.

* create: expose Inkling 2 Mini context length in model metadata

Copy model_max_length to max_position_embeddings during import so
ollama show reports the checkpoint's context length. Reject configs
that specify conflicting limits.

* mlxrunner: reduce Metal dispatches in Inkling 2 Mini decode

Reuse the prefill mask shader to compute relative position bias in one
dispatch for single-query attention. On M3 Ultra with Metal 3, compute
the loop controller's gain and forget projections in one matrix multiply.
2026-11-20 18:38:44 -08:00
..
architectures mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
audio llm: LFM3 GGUF fixes and llama-server speech (#19344) 2026-10-27 18:56:22 -07:00
autotrustdecision mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
clef mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
cohere2_moe mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
d1 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
d3 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
decisionqwen mlx: add d3 decision model support (#19196) 2026-10-22 08:50:14 -07:00
dflash mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
flint mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
gemma4 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
gemma4embedding mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
glimmer mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
glm4_moe_lite mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
inkling2 mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
kev mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
laguna mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
laya mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
lfm3 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
llama mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
nemotron_h mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
phonon mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
pplxdecider mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
qwen3 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
qwen3_5 mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
qwen3_5_moe mlxrunner, cmd, server: fold three small packages into their users 2026-09-16 14:06:08 -07:00
qwen4_exp mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
ripple mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
strands mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
embedding.go
embedding_test.go
globalscale.go New model... (#19030) 2026-10-15 07:28:26 -07:00
globalscale_test.go New model... (#19030) 2026-10-15 07:28:26 -07:00
linear.go
media.go mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
model.go mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
quant.go mlxrunner: lay out model by contract, checkpoint and construction 2026-09-16 14:06:08 -07:00
root.go mlxrunner: lay out model by contract, checkpoint and construction 2026-09-16 14:06:08 -07:00
score.go mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00
speech.go mlx: support concurrent generation (#19197) 2026-11-10 13:24:12 -08:00