---
license: other
base_model: Qwen/Qwen2.5-3B-Instruct
tags:
  - canxp
  - lora
  - peft
  - qlora
---

# vamman-2026-05-13-10-10

Fine-tuned by **CanXP AI** ([canxp.ai](https://canxp.ai)) from base model
`Qwen/Qwen2.5-3B-Instruct` using QLORA.

## Quick start (Python)

```bash
pip install transformers peft torch
```

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

base = "Qwen/Qwen2.5-3B-Instruct"
adapter = "canxp-ai/vamman-2026-05-13-10-10-ec369432"

tokenizer = AutoTokenizer.from_pretrained(base, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    base, torch_dtype="bfloat16", device_map="auto", trust_remote_code=True
)
model = PeftModel.from_pretrained(model, adapter)

prompt = "Hello!"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=200)
print(tokenizer.decode(out[0], skip_special_tokens=True))
```

## CLI download

```bash
pip install -U "huggingface_hub[cli]"
huggingface-cli download canxp-ai/vamman-2026-05-13-10-10-ec369432 --local-dir ./vamman-2026-05-13-10-10
```

## Training details

- Base model: `Qwen/Qwen2.5-3B-Instruct`
- Method: QLORA
- Epochs: 30
- Context length: 4096
- Validation split: 0.1

This adapter inherits the upstream license of the base model. See
LICENSE_NOTICE.txt in this repo for details.