CoolFace
Modelpublic

Saikrishna2511/java2py-qwen

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes102downloads
Model Card

Saikrishna2511/java2py-qwen

Java→Python fine-tuned Qwen2.5-Coder-0.5B-Instruct checkpoint (Stage 1 LoRA, merged for inference).

Demo

Related multi-task demo: https://huggingface.co/spaces/Saikrishna2511/qwen-multitask-demo

Task

Java → Python (java2py)

### Translate Java to Python:

{java code}

### Python:

## Training

- **Base model:** [Qwen/Qwen2.5-Coder-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B-Instruct)
- **Data:** AVATAR-TC / Java→Python pairs
- **Method:** LoRA (r=16, alpha=32), merged weights for inference

## Usage

from transformers import AutoModelForCausalLM, AutoTokenizer import torch

modelid = "Saikrishna2511/java2py-qwen" tokenizer = AutoTokenizer.frompretrained(modelid, trustremotecode=True) model = AutoModelForCausalLM.frompretrained( modelid, trustremotecode=True, torchdtype=torch.float16, device_map="auto", )

java = "public class Hello { public static void main(String[] args) { System.out.println(\"hi\"); } }" prompt = f"### Translate Java to Python:\\n``java\\n{java}\\n`\\n### Python:\\n``python\\n" inputs = tokenizer(prompt, returntensors="pt").to(model.device) outputs = model.generate(**inputs, maxnewtokens=512, temperature=0.2, topp=0.95) print(tokenizer.decode(outputs[0], skipspecialtokens=True))


## Limitations

- Small 0.5B model; translation quality varies with input complexity
- Prefer the multi-task checkpoint for NL→Python / Code2Doc: [Saikrishna2511/qwen-multitask](https://huggingface.co/Saikrishna2511/qwen-multitask)
- Not intended for production use without further evaluation