rayz8821/apartment-lora

TinyLlama-1.1B-Chat Fine-tuned with LoRA

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-Chat-v1.0 using LoRA (Low-Rank Adaptation).

Model Details

• Base Model: TinyLlama-1.1B-Chat-v1.0
• Fine-tuning Method: LoRA with QLoRA (4-bit quantization)
• LoRA Rank: 8
• LoRA Alpha: 32
• Target Modules: q_proj, v_proj
• Training Epochs: 5

Training Details

• Learning Rate: 2e-4
• Batch Size: 1 (with gradient accumulation of 16)
• Precision: FP16
• Final Training Loss: ~2.66
• Final Validation Loss: ~3.18

Usage

from peft import PeftModel
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load base model
base_model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load LoRA adapter
model = PeftModel.from_pretrained(model, "YOUR_USERNAME/apartment-lora")
tokenizer = AutoTokenizer.from_pretrained("YOUR_USERNAME/apartment-lora")

# Generate text
def generate_response(instruction, input_text=""):
    prompt = instruction
    if input_text:
        prompt += "\n" + input_text
    
    inputs = tokenizer(prompt, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_length=512,
            temperature=0.7,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )
    
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response[len(prompt):].strip()

# Example usage
response = generate_response("Explain quantum computing in simple terms")
print(response)

Training Data

The model was fine-tuned on custom QA datasets including:

• deepseek_qa.jsonl
• claude_qa.jsonl
• chat_qa.jsonl

Limitations

• This is a LoRA adapter, not a full model
• Requires the base model to function
• Performance depends on the quality of training data