metadata
language:
- en
license: apache-2.0
library_name: transformers
tags:
- code-generation
- python
- programming
- fine-tuned
pipeline_tag: text-generation
widget:
- text: |-
### Instruction:
Write a Python function to reverse a string
### Input:
### Response:
example_title: Reverse String
- text: |-
### Instruction:
how are you?
### Input:
### Response:
example_title: Non-Code Request
- text: |-
### Instruction:
Implement quicksort algorithm in Python
### Input:
Use recursion and list comprehensions
### Response:
example_title: Quicksort Algorithm
base_model: Mercy-62/python-code-only-Qwen-lora
datasets:
- sahil2801/CodeAlpaca-20k
- google-research-datasets/mbpp
- openai/openai_humaneval
🐍 Python Code-Only Qwen
A fine-tuned model that generates ONLY executable Python code, with no explanations or conversations.
🚀 Quick Inference Example
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load model
model = AutoModelForCausalLM.from_pretrained("Mercy-62/python-code-only-Qwen-lora")
tokenizer = AutoTokenizer.from_pretrained("Mercy-62/python-code-only-Qwen-lora")
# Use exact same prompt format from training
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
### Instruction:
{}
### Input:
{}
### Response:
{}"""
# Enable faster inference (if using Unsloth)
# FastLanguageModel.for_inference(model)
# Test 1: Simple code generation
inputs = tokenizer(
[
alpaca_prompt.format(
"Write a Python function to reverse a string", # instruction
"", # input (leave empty if no context)
"", # output - leave blank for generation
)
],
return_tensors="pt"
).to("cuda")
outputs = model.generate(**inputs, max_new_tokens=128, use_cache=True)
print("Test 1 - Reverse string function:")
print(tokenizer.batch_decode(outputs)[0])