| --- |
| language: |
| - en |
| - zh |
| license: apache-2.0 |
| base_model: |
| - khtsly/Mini-Coder-v2-noft |
| - Qwen/Qwen3.5-9B |
| - Qwen/Qwen3.5-9B-Base |
| tags: |
| - text-generation-inference |
| - transformers |
| - unsloth |
| - qwen |
| - qwen3.5 |
| - reasoning |
| - chain-of-thought |
| - lora |
| - luau |
| - roblox |
| pipeline_tag: image-text-to-text |
| datasets: |
| - khtsly/luau-stack-hq |
| - khtsly/luau-repo-docs-text |
| - Pinkstack/luaucoder-instructions-v3-SFT |
| --- |
| |
| >[!Note] |
| >We found loading with quantized like `Q4`, `Q3`, `Q2`, .. has performance degradation and repetitions. |
|
|
| # Mini-Coder-v2 (15B) |
|
|
| Mini-Coder-v2 is build on top of upscaled & merged of Qwen3.5-9B + Qwen3.5-9B-Base model with Continual Pretraining (CPT), we feed ~36.63k high-quality curated luau raw codes & docs text to improves the luau coding tasks & knowledges capability. |
|
|
| This model present a upscaled of Qwen3.5-9B from only 32 layers to 56 layers for a deeper reasoning capability! This model parameters is 15.07B with Vision, 14.60B without Vision. |
|
|
| # Uploaded finetuned model |
|
|
| - **Developed by:** khtsly |
| - **License:** apache-2.0 |
| - **Finetuned from model :** khtsly/Mini-Coder-v2-noft |
|
|
| This qwen3_5 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. |
| |
| [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth) |
