mirror/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2026-01-04 17:50:13 -06:00
Code Issues Packages Projects Releases Wiki Activity

chore(model gallery): 🤖 add 1 new models via gallery agent (#6854)

Browse Source 
chore(model gallery): 🤖 add new models via gallery agent

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: mudler <2420543+mudler@users.noreply.github.com>
This commit is contained in:
LocalAI [bot]
2025-10-28 09:23:07 +01:00
committed by GitHub
parent d95d4992fe
commit a946cb08b5
1 changed files with 38 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
gallery/index.yaml
View File

@@ -22721,3 +22721,41 @@
- filename: ChemDFM-R-14B.i1-Q4_K_M.gguf
sha256: c5090dd547154712865fa26389790cad93f23a7b5cba90f5efc8b667ad61cd06
uri: huggingface://mradermacher/ChemDFM-R-14B-i1-GGUF/ChemDFM-R-14B.i1-Q4_K_M.gguf
- !!merge <<: *qwen3
name: "qwen3-deckard-large-almost-human-6b-iii-160-omega"
urls:
- https://huggingface.co/mradermacher/Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA-GGUF
description: |
**Model Name:** Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA
**Base Model:** Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x
**Repository:** [DavidAU/Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA](https://huggingface.co/DavidAU/Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA)
**Description:**
A highly refined, large-scale fine-tuned version of Qwen3-6B, trained on an in-house dataset inspired by the works of Philip K. Dick. This model is part of the "Deckard" series, emphasizing deep reasoning, creative narrative, and human-like prose. Leveraging the innovative *Brainstorm 20x* training process, it enhances conceptual depth, coherence, and emotional engagement while maintaining strong instruction-following capabilities.
Optimized for long-context tasks (up to 256k tokens), it excels in code generation, creative writing, brainstorming, and complex reasoning. The model features a "heavy" fine-tuning (13% of parameters trained, 2x training duration) and includes an additional dataset of biographical and personal writings to restore narrative depth and authenticity.
**Key Features:**
- Trained using the *Brainstorm 20x* method for enhanced reasoning and narrative quality
- Supports 256k context length
- Ideal for creative writing, code generation, and step-by-step problem solving
- Fully compatible with GGUF, GPTQ, EXL2, AWQ, and HQQ formats
- Requires Jinja or CHATML template
**Use Case Highlights:**
- Long-form storytelling & worldbuilding
- Advanced coding with detailed reasoning
- Thoughtful brainstorming and idea development
- Roleplay and narrative-driven interaction
**Note:** The quantized version by mradermacher (e.g., `Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA-GGUF`) is derived from this source. For the full, unquantized model and best performance, use the original repository.
**License:** Apache 2.0
**Tags:** #Qwen3 #CodeGeneration #CreativeWriting #Brainstorm20x #PhilipKDick #LongContext #LLM #FineTuned #InstructModel
overrides:
parameters:
model: Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA.Q4_K_M.gguf
files:
- filename: Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA.Q4_K_M.gguf
sha256: c6c9c03e771edfb68d5eab82a3324e264e53cf1bcf9b80ae3f04bc94f57b1d7f
uri: huggingface://mradermacher/Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA-GGUF/Qwen3-Deckard-Large-Almost-Human-6B-III-160-OMEGA.Q4_K_M.gguf