Model Summary:
This model is a QLoRA fine-tuned adapter built on top of Meta-Llama-3.1-8B, trained to predict appliance prices from Amazon-style product descriptions.
The goal of this project is to produce grounded, realistic price estimates for appliances while keeping training lightweight enough to run on a free Google Colab T4 GPU.
The model outputs a dollar price when prompted with a product description ending in "Price is $".
Evaluation:
The fine-tuned adapter was evaluated against the base Frontier model using appliance pricing prompts.
Results:
The fine-tuned model produces more realistic and grounded price predictions
Outputs are less arbitrary and better aligned with real-world appliance pricing ranges
Especially strong on common appliances such as dishwashers, refrigerators, and microwaves
How to Use
Load the Model
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained(
"SameehK/Amazon-appliances-price-prediction-finetuned-model-2025-10-11_12.32.04",
device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(
"SameehK/Amazon-appliances-price-prediction-finetuned-model-2025-10-11_12.32.04"
)
Generate a Prediction
prompt = "Energy Star certified front-load washing machine with 4.5 cu ft capacity. Price is $"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=10)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
