This is a GGUF quantized version of TinyLlama (1.1B parameters) fine-tuned using Unsloth (2x faster training, 50% less memory) with LoRA adapters. The model has been optimized for efficient CPU inference with minimal memory footprint.

Key Features:

GGUF Format: Optimized for llama.cpp and CPU inference
Unsloth Optimized: 2x faster training, 30-50% less memory usage
Quantized: Reduced memory footprint without quality loss
Fine-tuned: Custom trained with Unsloth's optimized kernels

Model Details:

Base Model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
Fine-tuning Method: LoRA with Unsloth optimizations
Format: GGUF (GGML Universal Format)
Quantization: Q4_K_M (4-bit quantization)
Parameters: 1.1 Billion
Context Length: 2048 tokens

🚀 Usage

Option 1: Using llama-cpp-python (Recommended)

python

# Install llama-cpp-python
# pip install llama-cpp-python

from llama_cpp import Llama

# Load the model
model_path = "arif-butt/tinyllama-unsloth-gguf"  # or local path to .gguf file
llm = Llama(
    model_path=model_path,
    n_ctx=2048,              # Context length
    n_threads=4,             # Number of CPU threads
    n_gpu_layers=0,          # Set >0 for GPU offloading
    verbose=False,
)

# Simple prompt
prompt = "Q: Name all the courses Arif butt teach?\nA:"

# Generate response
output = llm(
    prompt,
    max_tokens=100,          # Maximum tokens to generate
    temperature=0.2,         # Lower = more deterministic
    top_p=0.95,              # Nucleus sampling
    repeat_penalty=1.1,      # Penalize repetition
    stop=["Q:", "\nQ:"],     # Stop sequences
)

print(output["choices"][0]["text"])
Option 2: Using Transformers with llama-cpp
from transformers import AutoTokenizer
from llama_cpp import Llama

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")

# Load GGUF model
model_path = "arif-butt/tinyllama-unsloth-gguf"
llm = Llama(model_path=model_path, n_ctx=2048)

def generate_response(prompt, max_tokens=100, temperature=0.7):
    """Generate response using the GGUF model"""
    output = llm(
        prompt,
        max_tokens=max_tokens,
        temperature=temperature,
        stop=["Q:", "\nQ:", "User:", "\nUser:"],
    )
    return output["choices"][0]["text"]

# Test
prompt = "Q: What is machine learning?\nA:"
response = generate_response(prompt)
print(f"Response: {response}")

Option 3: Using llama.cpp CLI
# Download the model
wget https://huggingface.co/arif-butt/tinyllama-unsloth-gguf/resolve/main/tinyllama-unsloth-q4_k_m.gguf

# Run inference
./main -m tinyllama-unsloth-q4_k_m.gguf \
       -p "Q: Name all the courses Arif butt teach?\nA:" \
       -n 100 \
       -t 4 \
       --temp 0.2 \
       --top_p 0.95
LORA_R       = 16        # Rank of LoRA matrices
LORA_ALPHA   = 32        # Scaling factor (alpha/r = 2.0)
LORA_DROPOUT = 0.05      # Dropout for regularization
TARGET_MODULES = [       # Layers where LoRA is applied
    "q_proj",            # Query projection
    "k_proj",            # Key projection  
    "v_proj",            # Value projection
    "o_proj",            # Output projection
    "gate_proj",         # Gate projection (MLP)
    "up_proj",           # Up projection (MLP)
    "down_proj"          # Down projection (MLP)
]