You can deploy this model and start earning money today!

dhanyaamanoj/
science-tutor-llama-3.2-1b-lora

Name: science-tutor-llama-3.2-1b-lora
Availability: InStock

safetensors

Views

No views yet

Science Tutor LLM (Llama 3.2 1B LoRA)

Model Description

This model is a fine-tuned version of Llama 3.2 1B Instruct designed to act as an AI Science Tutor.

It is optimized for:

Step-by-step explanations
Concept teaching (not just answering)
Structured responses (Goal → Steps → Example → Checkpoint)

Intended Use

This model is designed for:

Physics, Chemistry, Biology tutoring
Concept explanation for students
Guided learning and reasoning

Out-of-Scope Use

Medical or legal advice
Financial decision-making
Harmful or unsafe content

Training Details

Base Model

meta-llama/Llama-3.2-1B-Instruct

Fine-Tuning Method

QLoRA (4-bit quantization)
PEFT (Parameter Efficient Fine-Tuning)

Dataset

Synthetic tutoring dataset (~300–800 samples)
Format: Instruction → Structured Tutor Response

Features:

Step-by-step explanations
Student misconceptions
Teaching-style responses

⚙️ Training Hyperparameters

Epochs: 3
Batch size: 2 (effective 8 with grad accumulation)
Learning rate: 2e-4
Max sequence length: 512
Optimizer: paged_adamw_32bit
Precision: FP16

📊 Evaluation

Evaluation performed on 30+ science questions.

Metrics:

Correctness
Step-by-step reasoning
Clarity
Teaching quality
Structure

Results:

Metric	Base Model	Fine-Tuned
Correctness	3.8	4.5
Step-by-step	2.9	4.6
Clarity	3.5	4.4
Teaching Style	2.7	4.7
Structure	2.5	4.8
Overall	3.1	4.6

Fine-tuned model shows strong improvement in pedagogical quality and structured explanations.

Example

Question:

Why is the sky blue?

Response:

Goal: Understand why sky appears blue
Step 1: Sunlight contains multiple colors
Step 2: Atmosphere scatters light
Step 3: Blue scatters more than red
Example: Similar to dust scattering
Checkpoint: Why do sunsets appear red?

⚠️ Limitations

May still hallucinate in rare cases
Limited to general science topics
Not suitable for critical real-world decisions

Usage

python

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

model_id = "meta-llama/Llama-3.2-1B-Instruct"
adapter = "dhanyaamanoj/science-tutor-llama-3.2-1b-lora"

tokenizer = AutoTokenizer.from_pretrained(model_id)
base_model = AutoModelForCausalLM.from_pretrained(model_id)

model = PeftModel.from_pretrained(base_model, adapter)