From c7a903838cc31b1633e80d7d6f4768f5cf61b66f Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> Date: Mon, 12 Oct 2026 00:04:26 +0000 Subject: [PATCH] chore(gallery): add Swift 1.5 Q5 and Q6 builds Offer intermediate quantizations of Swift 1.5 Qwen3.8 27B alongside the existing Q4 and Q8 builds. Keep the F16 vision projector and publisher license notice, and pin the new downloads to one revision. Assisted-by: Codex:gpt-6 --- docs/content/features/model-gallery.md | 14 ++++ gallery/index.yaml | 98 ++++++++++++++++++++++++++ 2 files changed, 112 insertions(+) diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md index a5af4a904..550c3861c 100644 --- a/docs/content/features/model-gallery.md +++ b/docs/content/features/model-gallery.md @@ -205,6 +205,20 @@ The files use standard llama.cpp quantization types and the Apache-2.0 license. See the [publisher's model card](https://huggingface.co/agentionai/Qwen3.8-27B-AP-GGUF) for quantization details. These entries do not enable MTP speculative decoding. +## Swift 1.5 Qwen3.8-27B vision builds + +The `swift-1.5-qwen3.8-27b-q4` gallery entry offers Q4_K_M, Q5_K_M, Q6_K, +and Q8_0 variants for llama.cpp. Each build includes the F16 vision projector. +Select Q5_K_M or Q6_K for a size between the Q4_K_M and Q8_0 builds: + +```sh +local-ai models install swift-1.5-qwen3.8-27b-q4 --variant swift-1.5-qwen3.8-27b-q5 +local-ai models install swift-1.5-qwen3.8-27b-q4 --variant swift-1.5-qwen3.8-27b-q6 +``` + +The [Swift Open License v1.0](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/LICENSE) +requires a separate license for commercial use above its revenue threshold. + ## Swift 1.5 Qwen3.8-27B GSQ-RCO Install `swift-1.5-qwen3.8-27b-gsq-rco` for text chat with llama.cpp. diff --git a/gallery/index.yaml b/gallery/index.yaml index 947b4be7d..65424dec1 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -65895,6 +65895,8 @@ uri: https://huggingface.co/RavichandranJ/Dolphin3-Cyber-8B-GGUF/resolve/145d66f096c8a99aad63be19ada1fbc46d3230bd/Dolphin3.0-Llama3.1-8B-abliterated.Q8_0.gguf - name: swift-1.5-qwen3.8-27b-q4 variants: + - model: swift-1.5-qwen3.8-27b-q5 + - model: swift-1.5-qwen3.8-27b-q6 - model: swift-1.5-qwen3.8-27b-q8 url: github:mudler/LocalAI/gallery/virtual.yaml@master urls: @@ -65943,6 +65945,102 @@ - filename: mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf uri: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF/resolve/a1614465cfa35d04d3e8575d713fa779662b5eab/mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf sha256: 10a24dc46eb801ad794886ef27ea1e43634d35388dfaf2e2374b7a8b67faf526 +- name: swift-1.5-qwen3.8-27b-q5 + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b + - https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF + license: swift-open-license-1.0 + description: | + Swift 1.5 is a 27B Qwen3.8 fine-tune for reasoning, coding, and agent tasks. + This Q5_K_M GGUF build uses llama.cpp, the embedded chat template, and an F16 vision projector for image input. + The Swift Open License v1.0 requires a separate license for commercial use above its revenue threshold. + tags: + - llm + - gguf + - cpu + - gpu + - vision + - multimodal + - reasoning + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + options: + - use_jinja:true + parameters: + model: Swift-1.5-Qwen3.8-27B-Q5_K_M.gguf + temperature: 1 + top_p: 0.95 + top_k: 20 + min_p: 0 + presence_penalty: 0 + repeat_penalty: 1 + template: + use_tokenizer_template: true + files: + - filename: Swift-1.5-Qwen3.8-27B-Q5_K_M.gguf + sha256: 4964843f816dafe77ce24a92c97f3bb094b218d752172710d9438af0e2562931 + uri: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF/resolve/14bfe4b42be4a925d98816db830155f476c605e7/Swift-1.5-Qwen3.8-27B-Q5_K_M.gguf + - filename: mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + uri: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF/resolve/14bfe4b42be4a925d98816db830155f476c605e7/mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + sha256: 10a24dc46eb801ad794886ef27ea1e43634d35388dfaf2e2374b7a8b67faf526 +- name: swift-1.5-qwen3.8-27b-q6 + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b + - https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF + license: swift-open-license-1.0 + description: | + Swift 1.5 is a 27B Qwen3.8 fine-tune for reasoning, coding, and agent tasks. + This Q6_K GGUF build uses llama.cpp, the embedded chat template, and an F16 vision projector for image input. + The Swift Open License v1.0 requires a separate license for commercial use above its revenue threshold. + tags: + - llm + - gguf + - cpu + - gpu + - vision + - multimodal + - reasoning + overrides: + backend: llama-cpp + context_size: 32768 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + options: + - use_jinja:true + parameters: + model: Swift-1.5-Qwen3.8-27B-Q6_K.gguf + temperature: 1 + top_p: 0.95 + top_k: 20 + min_p: 0 + presence_penalty: 0 + repeat_penalty: 1 + template: + use_tokenizer_template: true + files: + - filename: Swift-1.5-Qwen3.8-27B-Q6_K.gguf + sha256: cd2eb6597026ce4eb86100e830be24f01461f84b89d4708634c2ec3d43b67fa8 + uri: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF/resolve/14bfe4b42be4a925d98816db830155f476c605e7/Swift-1.5-Qwen3.8-27B-Q6_K.gguf + - filename: mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + uri: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF/resolve/14bfe4b42be4a925d98816db830155f476c605e7/mmproj-Swift-1.5-Qwen3.8-27B-F16.gguf + sha256: 10a24dc46eb801ad794886ef27ea1e43634d35388dfaf2e2374b7a8b67faf526 - name: swift-1.5-qwen3.8-27b-q8 url: github:mudler/LocalAI/gallery/virtual.yaml@master urls: