ollama/model/renderers
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
..
testdata mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
audio.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
audio_test.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
cogito.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
cogito_test.go preserve tool definition and call JSON ordering (#13525) 2026-01-05 18:03:36 -08:00
cohere.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
cohere_test.go models: add cohere2_moe (Command A / North) to the MLX engine (#16670) 2026-06-16 23:15:21 -07:00
d3_reference_test.go mlx: add d3 decision model support (#19196) 2026-10-22 08:50:14 -07:00
deepseek3.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
deepseek3_test.go preserve tool definition and call JSON ordering (#13525) 2026-01-05 18:03:36 -08:00
flint.go flint: fix replies with thinking disabled (#19216) 2026-10-22 11:38:27 -07:00
flint_json.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
flint_reference_test.go flint: fix replies with thinking disabled (#19216) 2026-10-22 11:38:27 -07:00
flint_test.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
functiongemma.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
functiongemma_test.go preserve tool definition and call JSON ordering (#13525) 2026-01-05 18:03:36 -08:00
gemma4.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
gemma4_audio.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
gemma4_reference_test.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
glimmer.go mlx: System one support (#18701) 2026-10-03 13:17:21 -07:00
glimmer_reference_test.go model/renderers: match Muse Glimmer reasoning template (#17732) 2026-08-13 14:35:50 -07:00
glimmer_test.go Release v0.32.7 (#17646) 2026-08-10 04:04:56 -07:00
glm46.go runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031) 2026-05-29 13:35:47 -07:00
glm46_test.go model: add Glm4MoeLiteForCausalLM architecture to support GLM-4.7-Flash (#13779) 2026-01-19 12:47:17 -08:00
glm47.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
glm47_test.go model: add Glm4MoeLiteForCausalLM architecture to support GLM-4.7-Flash (#13779) 2026-01-19 12:47:17 -08:00
glmocr.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
glmocr_test.go anthropic: Preserve Claude local image-path tool results in renderer-owned prompt formatting (#16047) 2026-05-12 00:02:17 -04:00
image_tags.go anthropic: Preserve Claude local image-path tool results in renderer-owned prompt formatting (#16047) 2026-05-12 00:02:17 -04:00
image_tags_test.go anthropic: Preserve Claude local image-path tool results in renderer-owned prompt formatting (#16047) 2026-05-12 00:02:17 -04:00
inkling2.go mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
inkling2_reference_test.go mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
inkling2_test.go mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
json.go decision: fix JSON state prompts for MLX decision models (#18974) 2026-10-13 12:43:30 -07:00
json_test.go decision: fix JSON state prompts for MLX decision models (#18974) 2026-10-13 12:43:30 -07:00
laguna.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
laguna_test.go lint: clean up current tree (#17456) 2026-07-29 15:33:28 -07:00
lfm2.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
lfm2_test.go model: update lfm2 parser/renderer for optional thinking (#16359) 2026-06-14 20:37:08 -07:00
lfm3.go llm: LFM3 GGUF fixes and llama-server speech (#19344) 2026-10-27 18:56:22 -07:00
lfm3_audio.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
lfm3_reference_test.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
lfm3_speech.go llm: LFM3 GGUF fixes and llama-server speech (#19344) 2026-10-27 18:56:22 -07:00
lfm3_speech_test.go llm: LFM3 GGUF fixes and llama-server speech (#19344) 2026-10-27 18:56:22 -07:00
lfm3_test.go mlx: add LFM3 and d1 models (#19333) 2026-10-27 08:47:18 -07:00
mimo.go model: add MiMo V2.6 renderer for Qwen3.5 (#19400) 2026-11-03 07:09:06 -08:00
mimo_reference_test.go model: add MiMo V2.6 renderer for Qwen3.5 (#19400) 2026-11-03 07:09:06 -08:00
nemotron3nano.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
nemotron3nano_reference_test.go nemotron_h: support the Nemotron 3.5 prompt layout (#17672) 2026-08-11 06:18:51 -07:00
nemotron3nano_test.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
nemotron35_reference_test.go New model... (#19030) 2026-10-15 07:28:26 -07:00
nemotron35_super_test.go New model... (#19030) 2026-10-15 07:28:26 -07:00
nemotron_cases_test.go mlx: implement Nemotron 3 Nano Omni (#17060) 2026-08-10 21:42:34 -07:00
nemotron_extra_reference_test.go mlx: implement Nemotron 3 Nano Omni (#17060) 2026-08-10 21:42:34 -07:00
olmo3.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
olmo3_test.go olmo3: fix flaky test (#13629) 2026-01-05 22:37:20 -08:00
olmo3_think.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
olmo3_think_test.go preserve tool definition and call JSON ordering (#13525) 2026-01-05 18:03:36 -08:00
ornith.go parser/renderer: add Ornith 9B renderer/parser support (#16920) 2026-06-25 23:18:47 -07:00
ornith_test.go parser/renderer: add Ornith 9B renderer/parser support (#16920) 2026-06-25 23:18:47 -07:00
qwen3coder.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
qwen3coder_test.go api: preserve unmodeled tool schema keywords (#19686) 2026-11-13 16:00:07 -08:00
qwen3vl.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
qwen3vl_nonthinking_test.go anthropic: Preserve Claude local image-path tool results in renderer-owned prompt formatting (#16047) 2026-05-12 00:02:17 -04:00
qwen3vl_thinking_test.go model: support for qwen3.5 architecture (#14378) 2026-02-24 20:08:05 -08:00
qwen35.go api: expose model thinking levels and defaults (#18473) 2026-09-18 14:10:53 -07:00
qwen35_test.go model: fix renderer and parser for qwen3.5 (#14605) 2026-03-03 20:58:29 -08:00
qwen38_test.go mlx: add d3 decision model support (#19196) 2026-10-22 08:50:14 -07:00
renderer.go mlxrunner: add Inkling 2 Mini (#19799) 2026-11-20 18:38:44 -08:00
renderer_test.go Update gemma renderers (#19241) 2026-10-23 07:12:11 -07:00
ripple.go Add ripple preview support to the MLX runner (#19592) 2026-11-07 16:36:53 -08:00
ripple_test.go Add ripple preview support to the MLX runner (#19592) 2026-11-07 16:36:53 -08:00
speech.go llm: LFM3 GGUF fixes and llama-server speech (#19344) 2026-10-27 18:56:22 -07:00
testhelpers_test.go model: add Glm4MoeLiteForCausalLM architecture to support GLM-4.7-Flash (#13779) 2026-01-19 12:47:17 -08:00
thinking.go Map disabled thinking to the models lowest supported level (#19684) 2026-11-12 15:03:07 -08:00
thinking_test.go Map disabled thinking to the models lowest supported level (#19684) 2026-11-12 15:03:07 -08:00