chore(model gallery): add gemma-3-12b-fornaxv.2-qat-cot (#5337)
Some checks failed
Explorer deployment / build-linux (push) Has been cancelled
GPU tests / ubuntu-latest (1.21.x) (push) Has been cancelled
generate and publish intel docker caches / generate_caches (intel/oneapi-basekit:2025.1.0-0-devel-ubuntu22.04, linux/amd64, ubuntu-latest) (push) Has been cancelled
build container images / hipblas-jobs (-aio-gpu-hipblas, rocm/dev-ubuntu-22.04:6.1, hipblas, true, ubuntu:22.04, extras, latest-gpu-hipblas, latest-aio-gpu-hipblas, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, -hipblas) (push) Has been cancelled
build container images / hipblas-jobs (rocm/dev-ubuntu-22.04:6.1, hipblas, true, ubuntu:22.04, core, latest-gpu-hipblas-core, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, false, -hipblas-core) (push) Has been cancelled
build container images / self-hosted-jobs (-aio-gpu-intel-f16, quay.io/go-skynet/intel-oneapi-base:latest, sycl_f16, true, ubuntu:22.04, extras, latest-gpu-intel-f16, latest-aio-gpu-intel-f16, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, -sycl-f16) (push) Has been cancelled
build container images / self-hosted-jobs (-aio-gpu-intel-f32, quay.io/go-skynet/intel-oneapi-base:latest, sycl_f32, true, ubuntu:22.04, extras, latest-gpu-intel-f32, latest-aio-gpu-intel-f32, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, -sycl-f32) (push) Has been cancelled
build container images / self-hosted-jobs (-aio-gpu-nvidia-cuda-11, ubuntu:22.04, cublas, 11, 7, true, extras, latest-gpu-nvidia-cuda-11, latest-aio-gpu-nvidia-cuda-11, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, -cublas-cuda11) (push) Has been cancelled
build container images / self-hosted-jobs (-aio-gpu-nvidia-cuda-12, ubuntu:22.04, cublas, 12, 0, true, extras, latest-gpu-nvidia-cuda-12, latest-aio-gpu-nvidia-cuda-12, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, -cublas-cuda12) (push) Has been cancelled
build container images / self-hosted-jobs (quay.io/go-skynet/intel-oneapi-base:latest, sycl_f16, true, ubuntu:22.04, core, latest-gpu-intel-f16-core, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, false, -sycl-f16-core) (push) Has been cancelled
build container images / self-hosted-jobs (quay.io/go-skynet/intel-oneapi-base:latest, sycl_f32, true, ubuntu:22.04, core, latest-gpu-intel-f32-core, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, false, -sycl-f32-core) (push) Has been cancelled
build container images / self-hosted-jobs (ubuntu:22.04, , true, extras, --jobs=3 --output-sync=target, linux/amd64, arc-runner-set, auto, ) (push) Has been cancelled
build container images / core-image-build (-aio-cpu, ubuntu:22.04, , true, core, latest-cpu, latest-aio-cpu, --jobs=4 --output-sync=target, linux/amd64,linux/arm64, arc-runner-set, false, auto, -core) (push) Has been cancelled
build container images / core-image-build (ubuntu:22.04, cublas, 11, 7, true, core, latest-gpu-nvidia-cuda-12-core, --jobs=4 --output-sync=target, linux/amd64, arc-runner-set, false, false, -cublas-cuda11-core) (push) Has been cancelled
build container images / core-image-build (ubuntu:22.04, cublas, 12, 0, true, core, latest-gpu-nvidia-cuda-12-core, --jobs=4 --output-sync=target, linux/amd64, arc-runner-set, false, false, -cublas-cuda12-core) (push) Has been cancelled
build container images / core-image-build (ubuntu:22.04, vulkan, true, core, latest-gpu-vulkan-core, --jobs=4 --output-sync=target, linux/amd64, arc-runner-set, false, false, -vulkan-core) (push) Has been cancelled
build container images / gh-runner (nvcr.io/nvidia/l4t-jetpack:r36.4.0, cublas, 12, 0, true, core, latest-nvidia-l4t-arm64-core, --jobs=4 --output-sync=target, linux/arm64, ubuntu-24.04-arm, true, false, -nvidia-l4t-arm64-core) (push) Has been cancelled
Security Scan / tests (push) Has been cancelled
Tests extras backends / tests-transformers (push) Has been cancelled
Tests extras backends / tests-rerankers (push) Has been cancelled
Tests extras backends / tests-diffusers (push) Has been cancelled
Tests extras backends / tests-coqui (push) Has been cancelled
tests / tests-linux (1.21.x) (push) Has been cancelled
tests / tests-aio-container (push) Has been cancelled
tests / tests-apple (1.21.x) (push) Has been cancelled

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2025-05-08 12:07:25 +02:00 committed by GitHub
parent de786f6586
commit d5e032bdcd
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -1287,6 +1287,32 @@
- filename: Comet_12B_V.5.i1-Q4_K_M.gguf
sha256: 02b5903653f1cf8337ffbd506b55398daa6e6e31474039ca4a5818b0850e3845
uri: huggingface://mradermacher/Comet_12B_V.5-i1-GGUF/Comet_12B_V.5.i1-Q4_K_M.gguf
- !!merge <<: *gemma3
name: "gemma-3-12b-fornaxv.2-qat-cot"
icon: https://huggingface.co/ConicCat/Gemma-3-12B-FornaxV.2-QAT-CoT/resolve/main/Fornax.jpg
urls:
- https://huggingface.co/ConicCat/Gemma-3-12B-FornaxV.2-QAT-CoT
- https://huggingface.co/mradermacher/Gemma-3-12B-FornaxV.2-QAT-CoT-GGUF
description: |
This model is an experiment to try to produce a strong smaller thinking model capable of fitting in an 8GiB consumer graphics card with generalizeable reasoning capabilities. Most other open source thinking models, especially on the smaller side, fail to generalize their reasoning to tasks other than coding or math due to an overly large focus on GRPO zero for CoT which is only applicable for coding and math.
Instead of using GRPO, this model aims to SFT a wide variety of high quality, diverse reasoning traces from Deepseek R1 onto Gemma 3 to force the model to learn to effectively generalize its reasoning capabilites to a large number of tasks as an extension of the LiMO paper's approach to Math/Coding CoT. A subset of V3 O3/24 non-thinking data was also included for improved creativity and to allow the model to retain it's non-thinking capabilites.
Training off the QAT checkpoint allows for this model to be used without a drop in quality at Q4_0, requiring only ~6GiB of memory.
Thinking Mode
Similar to the Qwen 3 model line, Gemma Fornax can be used with or without thinking mode enabled.
To enable thinking place /think in the system prompt and prefill <think>\n for thinking mode.
To disable thinking put /no_think in the system prompt.
overrides:
parameters:
model: Gemma-3-12B-FornaxV.2-QAT-CoT.Q4_K_M.gguf
files:
- filename: Gemma-3-12B-FornaxV.2-QAT-CoT.Q4_K_M.gguf
sha256: 75c66d64a32416cdaaeeeb1d11477481c93558ade4dc61a93f7aba8312cd0480
uri: huggingface://mradermacher/Gemma-3-12B-FornaxV.2-QAT-CoT-GGUF/Gemma-3-12B-FornaxV.2-QAT-CoT.Q4_K_M.gguf
- &llama4
url: "github:mudler/LocalAI/gallery/llama3.1-instruct.yaml@master"
icon: https://avatars.githubusercontent.com/u/153379578