Skip to content
Other

skillforge-skiller-0.5B

a compact language model that runs right in your browser

2 min read

This fine-tuned version of the compact Qwen2.5-0.5B-Instruct base model is optimized for turning raw source material into finished skill documents. It was trained using LoRA (a lightweight adapter on top of the base model) on material-to-skill pairs tailored to a specific system prompt. Quantized to the q4f16_1 format via mlc-llm, the model runs directly in the browser through WebLLM with no remote servers needed. It is a great fit for generating short, structured documents locally on your device.

Key features

  • Base model: Qwen2.5-0.5B-Instruct with an unchanged architecture.
  • Training: LoRA fine-tune (r=16) on training pairs matched to a strict document format.
  • Format: MLC q4f16_1 (8 shards, roughly 278 MB total).
  • Runtime: runs locally in the browser via WebLLM using WebGPU.

What you can do

  • Plug the model into a WebLLM-powered web app to generate skill documents locally in the browser.
  • Use it as a free, private on-device option without sending data to a server.
  • Generate short, cleanly structured skills and text assets in resource-constrained environments.

Example

import { CreateMLCEngine } from "@mlc-ai/web-llm";

const engine = await CreateMLCEngine("skillforge-skiller-0.5B", {
  appConfig: {
    model_list: [
      {
        model_id: "skillforge-skiller-0.5B",
        model: "[https://huggingface.co/Ciao22/skillforge-skiller-0.5B](https://huggingface.co/Ciao22/skillforge-skiller-0.5B)",
        model_lib: "[https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/web-llm-models/v0_2_87/base/Qwen2-0.5B-Instruct-q4f16_1_cs1k-webgpu.wasm](https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/web-llm-models/v0_2_87/base/Qwen2-0.5B-Instruct-q4f16_1_cs1k-webgpu.wasm)",
        vram_required_MB: 944.62,
        low_resource_required: true,
        overrides: { context_window_size: 4096 },
      },
    ],
  },
});
Read moreDetails
VRAM requirements: around 944.62 MB.
Limitations: trained on a small set of 12 handcrafted pairs and sitting at just 0.5B parameters, the model works best on short tasks and is not a replacement for large flagship models.
Discuss Telegram 1

Will it run?

Runs in GGUF Q4_K_M

The most precise of the ready quantised files fits.

  • GGUF Q4_K_Mweights 0.3 GB · ready files0.8 GB

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: GGUF Q4_K_M — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Ciao22/skillforge-skiller-0.5B
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Ciao22/skillforge-skiller-0.5B --local-dir ~/models/skillforge-skiller-0.5B
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen2.5-0.5Bfamily root
  2. Qwen2.5-0.5B-Instruct
  3. skillforge-skiller-0.5B

New posts about versions of Qwen2.5-0.5B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 5All on Hugging Face
Merges 2All on Hugging Face

About the model

Ciao22/skillforge-skiller-0.5B
Author
Ciao22
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments