From f38d1758759c19d420ea67a0491fb9b68f21038c Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> Date: Wed, 14 Oct 2026 20:05:04 +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 --- docs/content/features/model-gallery.md | 20 ++++ gallery/index.yaml | 126 +++++++++++++++++++++ gallery/models/index.yaml | 150 +++++++++++++++++++++++++ 3 files changed, 296 insertions(+) diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md index 56dde2118..65496282b 100644 --- a/docs/content/features/model-gallery.md +++ b/docs/content/features/model-gallery.md @@ -194,6 +194,26 @@ To select a specific quantization, use `--variant` with `lfm2.5-24b-a2b` 131,072-token context, and LiquidAI's recommended sampling settings. The weights use the [Liquid AI LFM Open License](https://huggingface.co/LiquidAI/LFM2.5-24B-A2B/blob/main/LICENSE). +## AtomicChat LFM2.5-24B-A2B Uncensored + +[AtomicChat's LFM2.5-24B-A2B derivative](https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF) +reduces refusal behavior in LiquidAI's 24B mixture-of-experts model. The gallery +provides AD-Q4_K_M, AD-Q6_K, and AD-Q8_0 GGUF builds through `llama-cpp`. +The edits are included in each build; no separate LoRA adapter is needed. + +Install with automatic variant selection: + +```bash +local-ai models install atomic-lfm2.5-24b-a2b-uncensored +``` + +Use `--variant atomic-lfm2.5-24b-a2b-uncensored` to select AD-Q4_K_M, +or append `-q6` or `-q8` to select a higher-precision build. These entries +use the embedded chat template, a 32,768-token context, temperature 0.1, +top-k 50, and repetition penalty 1.05. The model supports contexts up to +131,072 tokens. The weights use the +[Liquid AI LFM Open License](https://huggingface.co/AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF/blob/main/LICENSE). + ## Cyber-Ornith 1.5 9B Cyber-Ornith 1.5 is a Qwen3.5 fine-tune for security auditing, terminal tasks, and tool use. diff --git a/gallery/index.yaml b/gallery/index.yaml index 981cc3563..13fe9b628 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: adavar-3b-q4 description: | diff --git a/gallery/models/index.yaml b/gallery/models/index.yaml index da124ab3a..e81655051 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: adavar-3b-q4 name: adavar-3b-q4