codeparrot/code-generator

1import gradio as gr2from transformers import AutoTokenizer, AutoModelForCausalLM, set_seed, pipeline3 4 5title = "Code Generator"6description = "This is a space to convert english text to Python code using with [codeparrot-small-text-to-code](https://huggingface.co/codeparrot/codeparrot-small-text-to-code),\7            a code generation model for Python finetuned on [github-jupyter-text](https://huggingface.co/datasets/codeparrot/github-jupyter-text) a dataset of doctrings\8            and their Python code extracted from Jupyter notebooks."9example = [10    ["Utility function to compute the accuracy of predictions using metric from sklearn", 65, 0.6, 42],11    ["Let's implement a function that computes the size of a file called filepath", 60, 0.6, 42],12    ["Let's implement bubble sort in a helper function:", 87, 0.6, 42],13    ]14 15# change model to the finetuned one16tokenizer = AutoTokenizer.from_pretrained("codeparrot/codeparrot-small-text-to-code")17model = AutoModelForCausalLM.from_pretrained("codeparrot/codeparrot-small-text-to-code")18 19def make_doctring(gen_prompt):20    return "\"\"\"\n" + gen_prompt + "\n\"\"\"\n\n"21 22def code_generation(gen_prompt, max_tokens, temperature=0.6, seed=42):23    set_seed(seed)24    pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)25    prompt = make_doctring(gen_prompt)26    generated_text = pipe(prompt, do_sample=True, top_p=0.95, temperature=temperature, max_new_tokens=max_tokens)[0]['generated_text']27    return generated_text28 29 30iface = gr.Interface(31    fn=code_generation, 32    inputs=[33        gr.Code(lines=10, language="python", label="English instructions"),34        gr.inputs.Slider(35            minimum=8,36            maximum=256,37            step=1,38            default=8,39            label="Number of tokens to generate",40        ),41        gr.inputs.Slider(42            minimum=0,43            maximum=2.5,44            step=0.1,45            default=0.6,46            label="Temperature",47        ),48        gr.inputs.Slider(49            minimum=0,50            maximum=1000,51            step=1,52            default=42,53            label="Random seed to use for the generation"54        )55    ],56    outputs=gr.Code(label="Predicted Python code", language="python", lines=10),57    examples=example,58    layout="horizontal",59    theme="peach",60    description=description,61    title=title62)63iface.launch()64