Modelpublic

izambasiron/qwen3-0.6b-markdown-review

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face

0likes57downloads

Model Card

qwen3-0.6b-markdown-review

Fine-tuned Qwen3 0.6B for automated Markdown document review — catches missing alt text, broken links, stale filenames, and formatting issues.

Model details


Base model	`unsloth/Qwen3-0.6B`
Parameters	0.6B
Fine-tuning	QLoRA (4-bit)
Training data	168 curated examples
Eval data	42 held-out examples
Epochs	3
Precision	fp16
VRAM (training)	~6 GB (T4)

Usage

transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "izambasiron/qwen3-0.6b-markdown-review",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("izambasiron/qwen3-0.6b-markdown-review")

messages = [
    {"role": "system", "content": "You are a Markdown document reviewer..."},
    {"role": "user", "content": "# My Doc\n\n![img](./old.png)"}
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Ollama / llama.cpp

Download qwen3-0.6b.Q4_K_M.gguf and use with the included Modelfile:

bash

ollama create qwen3-markdown-review -f Modelfile
ollama run qwen3-markdown-review

Or directly with llama.cpp:

bash

llama-cli -m qwen3-0.6b.Q4_K_M.gguf -p "<system prompt>" -f input.md

Limitations

—Trained on 168 examples — may overfit
—English only
—Not suitable for general chat — task-specific fine-tune
—Requires ~400 MB RAM for Q4KM GGUF; ~1.4 GB for fp16 transformers

Training

Fine-tuned with Unsloth using the notebook at `colab_finetune_qwen3_clean.ipynb`.