Fine-tuned Meta's Llama-3.2-3B-Instruct on agricultural question-answering using QLoRA via Unsloth — enabling accurate, domain-specific responses to farming and agriculture queries.

Model Details

Base Model: meta-llama/Llama-3.2-3B-Instruct
Fine-Tuning Method: SFT (Supervised Fine-Tuning) with QLoRA
Quantization: 4-bit
LoRA Targets: Attention + MLP layers
Library: Unsloth + TRL + PEFT

Dataset

Dataset: KisanVaani/agriculture-qa-english-only (HuggingFace)
English-only agricultural Q&A pairs formatted using Llama-3.2 chat template

Training Config

Parameter	Value
Batch Size	2
Gradient Accumulation	4
Learning Rate	2e-4
Epochs	1
Quantization	4-bit

How to Use

python

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="your-username/your-repo-name",
    load_in_4bit=True,
)

FastLanguageModel.for_inference(model)

messages = [{"role": "user", "content": "How do I treat blight in tomatoes?"}]
inputs = tokenizer.apply_chat_template(
    messages, tokenize=True, add_generation_prompt=True, return_tensors="pt"
).to("cuda")

outputs = model.generate(input_ids=inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Limitations

Trained for 1 epoch on English agricultural data only
Best suited for general farming Q&A, not highly localized advice

Uploaded model

Developed by: gajju17
License: apache-2.0
Finetuned from model : unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit

This llama model was trained 2x faster with Unsloth

gajju17/agri_lora_model

Llama-3.2-3B Agricultural Q&A Fine-Tune