from unsloth import FastLanguageModel
import torch

# Load base model
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="unsloth/Meta-Llama-3.1-8B-Instruct",
    max_seq_length=2048,
    dtype=None,
    load_in_4bit=True,
    device_map="auto",  # Use this if you have memory issues
)

# Load this adapter
model.load_adapter("menikev/Llama-3.1-8B-NigerianLegalBot", adapter_name="default")

# Enable inference
FastLanguageModel.for_inference(model)

# Use the model
system_prompt = "You are a helpful and knowledgeable legal assistant in Nigeria. Respond to queries in the language of the user's question."

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "What are the requirements for business registration in Nigeria?"}
]

inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")

with torch.no_grad():
    outputs = model.generate(
        input_ids=inputs.to("cuda"),
        max_new_tokens=512,
        temperature=0.7,
        do_sample=True,
    )

response = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
print(response)

Training Details

Base Model: unsloth/Meta-Llama-3.1-8B-Instruct
Method: QLoRA (4-bit quantization)
Rank: 16
Alpha: 16
Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Training Data: Custom Nigerian legal Q&A dataset in multiple languages

Files Uploaded

README.md
adapter_model.safetensors
adapter_config.json

Disclaimer

This model is for informational purposes only. Always consult qualified legal professionals for actual legal advice.

License

This adapter follows the LLaMA 3.1 license terms.