Catalagile-5

<p align="center"> <img src="https://huggingface.co/sandylolpotty/Catalagile-5/resolve/main/logo.png" alt="Model Logo" width="200"/> </p> <h1 align="center">🧠 Catalagile-5 by Aryamann Patne</h1> <h3 align="center">Fine-tuned Conversational Model using LoRA</h3>

---

🧾 Model Summary

This is a LoRA fine-tuned version of Microsoft’s DialoGPT-medium, optimized for smooth, natural, and contextually aware conversations.
It learns from a curated question-answer dataset to produce coherent and personalized dialogue — perfect for lightweight chatbots and assistants.

---

📘 Model Details

• Developer: Aryamann Patne
• Model Type: Large Language Model
• Base Model: microsoft/DialoGPT-medium
• Fine-tuning Framework: PEFT (Parameter-Efficient Fine-Tuning)
• Languages: English
• License: MIT
• Training Environment: Local (GPU A100)
• Fine-tuned Dataset: Custom QA dataset (``)

---

🔗 Model Sources

• Repository: https://huggingface.co/sandylolpotty/Catalagile-5
• Base Model: https://huggingface.co/microsoft/DialoGPT-medium
• Frameworks: PyTorch, Transformers, Datasets, PEFT

---

🎯 Intended Uses

Direct Use

• Chatbots
• Personal Assistants
• Roleplay / Story Bots
• Educational Question Answering

Downstream Fine-tuning

• Domain-specific assistants (e.g., legal, customer support, education)
• Emotionally adaptive or multilingual conversational bots

Out-of-Scope

• Disinformation or harmful generation
• Explicit or offensive outputs
• Any context requiring factual precision without validation

---

⚠️ Bias, Risks, and Limitations

While trained responsibly, this model may still:

• Reflect social or linguistic bias from its base data
• Produce hallucinated or incorrect facts
• Respond with verbosity or mild incoherence in edge cases

Recommendation: Always review model outputs before public deployment.

---

🚀 How to Use

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "AryamannPatne/DialoGPT-medium-LoRA"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

input_text = "Hey, how are you doing today?"
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_length=100, do_sample=True, top_p=0.9)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))