README.md

---
base_model: unsloth/Qwen3-Coder-30B-A3B-Instruct
base_model_relation: adapter
library_name: peft
license: apache-2.0
language:
- code
model_name: qwen3-coder-30b-a3b-codemonkey
pipeline_tag: text-generation
tags:
- lora
- peft
- qwen3
- qwen3-coder
- qwen3moe
- sft
- code
- unsloth
---

# qwen3-coder-30b-a3b-codemonkey

LoRA adapter for `unsloth/Qwen3-Coder-30B-A3B-Instruct`.

## Files

- `adapter_model.safetensors`: adapter weights
- `adapter_config.json`: PEFT config
- `tokenizer.json`, `tokenizer_config.json`, `chat_template.jinja`: tokenizer and chat template assets

## Load with Transformers + PEFT

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

base_id = "unsloth/Qwen3-Coder-30B-A3B-Instruct"
adapter_id = "1337Hero/qwen3-coder-30b-a3b-codemonkey"

tokenizer = AutoTokenizer.from_pretrained(base_id)
base_model = AutoModelForCausalLM.from_pretrained(
    base_id,
    torch_dtype="auto",
    device_map="auto",
)
model = PeftModel.from_pretrained(base_model, adapter_id)

messages = [
    {"role": "user", "content": "Write a Python function that atomically replaces a file."}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512)
completion = outputs[0][inputs.input_ids.shape[1]:]
print(tokenizer.decode(completion, skip_special_tokens=True))
```

## Adapter details

- Base model: `unsloth/Qwen3-Coder-30B-A3B-Instruct`
- PEFT type: `LoRA`
- Rank: `r=16`
- Alpha: `32`
- Target modules: `q_proj`, `k_proj`, `v_proj`, `o_proj`

## GGUF

A merged GGUF release can live in a separate repo such as
`1337Hero/qwen3-coder-30b-a3b-codemonkey-GGUF`.