From 1a6759a87308ba5e20f2eb39f1ed9ab2f8fd9468 Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> Date: Sun, 25 Oct 2026 09:51:56 +0000 Subject: [PATCH] feat(gallery): add Atomic LFM2.5 24B variants Add AtomicChat's LFM2.5-24B-A2B derivative with Q4, Q6, and Q8 GGUF builds for llama-cpp. Pin downloads and document installation. Assisted-by: Codex:gpt-6 Signed-off-by: Ettore Di Giacinto (cherry picked from commit f38d1758759c19d420ea67a0491fb9b68f21038c, mudler/LocalAI#12696) Assisted-by: Claude:claude-sonnet-5-5 Signed-off-by: Ettore Di Giacinto --- gallery/index.yaml | 126 ++++++++++++++++++++++++++++++++ gallery/models/index.yaml | 150 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 276 insertions(+) diff --git a/gallery/index.yaml b/gallery/index.yaml index 50581e4fa..26fbab071 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -3,6 +3,132 @@ # written by scripts/models/cmd/modelcatalog and stays only so releases before # the cutover keep working. CI fails when it differs from the generated copy: # go run ./scripts/models/cmd/modelcatalog --library gallery/models/index.yaml --out gallery/index.yaml +- url: github:mudler/LocalAI/gallery/virtual.yaml@master + name: atomic-lfm2.5-24b-a2b-uncensored + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q4_K_M GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + sha256: 17a5de7f9718d8f6133ef08579d35845dd69bb51f329204ead0fc0ee97a75a41 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true + variants: + - model: atomic-lfm2.5-24b-a2b-uncensored-q6 + - model: atomic-lfm2.5-24b-a2b-uncensored-q8 +- url: github:mudler/LocalAI/gallery/virtual.yaml@master + name: atomic-lfm2.5-24b-a2b-uncensored-q6 + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q6_K GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + sha256: 2eea8a0f214304357f45527b2bbc757a810445e8cbb38d7664115ac426aa0711 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true +- url: github:mudler/LocalAI/gallery/virtual.yaml@master + name: atomic-lfm2.5-24b-a2b-uncensored-q8 + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q8_0 GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + sha256: 852b43bc93dfc776f702c9ef0661c1e8cf92eb2d6914def912c51a2a9d45a4d1 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true - url: github:mudler/LocalAI/gallery/virtual.yaml@master name: atomic-minicpm-v-4.7-35b-a3b-uncensored-q4 description: | diff --git a/gallery/models/index.yaml b/gallery/models/index.yaml index b5438652c..1e78a5920 100644 --- a/gallery/models/index.yaml +++ b/gallery/models/index.yaml @@ -4,6 +4,156 @@ # changes and never edit it by hand. CI fails when it is stale. A build from # master reads it as the default model source; a tagged build reads the gallery # until the release cutover. +- kind: model + id: atomic-lfm2.5-24b-a2b-uncensored + name: atomic-lfm2.5-24b-a2b-uncensored + version: 0.0.0-gallery.83179d51480f + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q4_K_M GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + channel: gallery + policy: + requireSignature: false + schemaVersion: 1 + model: + url: github:mudler/LocalAI/gallery/virtual.yaml@master + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + sha256: 17a5de7f9718d8f6133ef08579d35845dd69bb51f329204ead0fc0ee97a75a41 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q4_K_M.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true + variants: + - model: atomic-lfm2.5-24b-a2b-uncensored-q6 + - model: atomic-lfm2.5-24b-a2b-uncensored-q8 +- kind: model + id: atomic-lfm2.5-24b-a2b-uncensored-q6 + name: atomic-lfm2.5-24b-a2b-uncensored-q6 + version: 0.0.0-gallery.810d2644dbd4 + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q6_K GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + channel: gallery + policy: + requireSignature: false + schemaVersion: 1 + model: + url: github:mudler/LocalAI/gallery/virtual.yaml@master + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + sha256: 2eea8a0f214304357f45527b2bbc757a810445e8cbb38d7664115ac426aa0711 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q6_K.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true +- kind: model + id: atomic-lfm2.5-24b-a2b-uncensored-q8 + name: atomic-lfm2.5-24b-a2b-uncensored-q8 + version: 0.0.0-gallery.8ea7aed2c62f + description: | + AtomicChat's LFM2.5-24B-A2B derivative reduces refusal behavior in the + original weights. This text-only mixture-of-experts model has 24B total + parameters and supports multilingual reasoning and tool use. + This entry uses the AD-Q8_0 GGUF build with the edits baked in. + license: other + urls: + - https://huggingface.co/LiquidAI/LFM2.5-24B-A2B + - https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - uncensored + channel: gallery + policy: + requireSignature: false + schemaVersion: 1 + model: + url: github:mudler/LocalAI/gallery/virtual.yaml@master + files: + - filename: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + sha256: 852b43bc93dfc776f702c9ef0661c1e8cf92eb2d6914def912c51a2a9d45a4d1 + uri: https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/resolve/1acb8f6eb027d85b7771dcc9eb68817aa480e9c8/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - completion + options: + - use_jinja:true + parameters: + model: llama-cpp/models/atomic-lfm2.5-24b-a2b-uncensored/LFM2.5-24B-A2B-Uncensored-AD-Q8_0.gguf + repeat_penalty: 1.05 + temperature: 0.1 + top_k: 50 + template: + use_tokenizer_template: true - kind: model id: atomic-minicpm-v-4.7-35b-a3b-uncensored-q4 name: atomic-minicpm-v-4.7-35b-a3b-uncensored-q4