🧠 Phi-2 Humanizer (LoRA Fine-Tuned)

This model is a fine-tuned version of microsoft/phi-2 on a dataset of 50+ examples. It specializes in making stiff or robotic AI-generated writing sound more natural, emotional, and conversational — while avoiding em/en dashes and overly formal phrasing.

✨ Demo

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("judesv/phi2-svHUCustxt")
tokenizer = AutoTokenizer.from_pretrained("judesv/phi2-svHUCustxt")

def humanize(prompt):
    full_prompt = (
        f"Instruct: Humanize the following AI-generated text.\n"
        f"Input: {prompt}\n"
        f"Output:"
    )
    inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(
        **inputs,
        max_new_tokens=100,
        do_sample=True,
        top_p=0.9,
        temperature=0.7,
        eos_token_id=tokenizer.eos_token_id
    )
    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return result[len(full_prompt):].strip()

gr.Interface(fn=humanize, inputs="text", outputs="text", title="🧠 Phi-2 Humanizer").launch()

Downloads last month: -; Downloads are not tracked for this model. How to track