ZainYasir/Puck-Perosnalized-bot

Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->

🧠 Puck Peronalized Bot

Model Details

A TinyLlama-based personalized conversational model trained on 5,000+ samples of English and Roman Urdu messages by Zain Yasir, reflecting his unique tone, knowledge, beliefs, and friend circle. Designed to power a private AI assistant named Puck.

Model Description

<!-- Provide a longer summary of what this model is. --> 🧩 Model Description This is a 1.1B-parameter TinyLlama model fine-tuned using LoRA (4-bit) on personal, technical, religious, and conversational data. It understands (English) text and is tailored to mimic natural, reflective, and casual conversations based on the user’s own messaging history.

• —License: MIT
• —**Finetuned from model : 5,000+ messages, custom instruction-response format

Model Sources [optional]

<!-- Provide the basic links for the model. -->

• —Repository: [More Information Needed]
• —Paper [optional]: [More Information Needed]
• —Demo [optional]: [More Information Needed]

Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

Direct Use

• —Chatbot for personal productivity, task planning, and faith-aligned reminders.
• —Assisting in small talk, Q&A, and self-reflective prompts.
• —Custom assistants (e.g., Puck on local apps or APIs).

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->

Downstream Use [optional]

• —Can be extended with RAG for dynamic factual recall.
• —Useful as a base for personalized LLM agents or lightweight voice assistants.

<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

• —Not for production-scale systems (use larger models instead).
• —Not suitable for sensitive decision-making or medical/legal advice.

Training Details

📚 Training Data

• —The model was trained on a curated dataset including:
• —600+ facts about Zain and friends (Q&A format × paraphrased)
• —500+ general conversations (e.g., daily routine, habits)
• —200+ tech/personal Q&A (projects, skills, tools)
• —3,700+ random Roman Urdu + English chats (faith, Pakistan, jokes, thoughts)

⚙️ Hyperparameters

• —Epochs: 3
• —Batch size: 4 × 4 (with gradient accumulation)
• —LR: 2e-4
• —Precision: FP16
• —LoRA config: r=8, alpha=16, target: qproj, vproj

Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• —Hardware Type: 2× NVIDIA T4
• —Hours used: ~2.5
• —Cloud Provider: Kaggle (Google Cloud)
• —Compute Region: Pakistan
• —Carbon Emitted: ~0.25 kg CO2e