Llama-3-TKK-8B-Elite-V1.0, a generative model built upon the LLaMA 8B architecture, represents my individual undergraduate graduation project. Developed during my studies in Software Engineering at Malatya Turgut Özal University, this project stands as a culmination of my academic endeavors. I extend my sincere appreciation to Assoc. Prof. Dr. Harun Bingöl, who served as both my department chair and thesis advisor. His invaluable guidance, unwavering support, and mentorship have significantly shaped my educational and research experiences. I am deeply grateful for his continuous encouragement, insightful feedback, and unwavering dedication. Thank you, Dr. Bingöl...

Model Details

Training took 133 hours and 59 minutes for a total of 37,420 steps and was conducted on 8 Tesla V100 GPUs.

Base Model: LLaMA 8B based LLM
Model Developers: Tarık Kaan Koç
Thesis Advisor: Assoc. Prof. Dr. Harun Bingöl
Input: Text only
Output: Text only
Training Dataset: Cleaned Turkish raw data with 1 million raw instruction Turkish data, private
Training Method: Fine-tuning with LORA

LORA Fine-Tuning Configuration

lora_alpha: 16
lora_dropout: 0.1
r: 64
bias: none
task_type: CAUSAL_LM

Example Usage:

python

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer, pipeline
import torch

model_id = "tarikkaankoc7/TKK-LLaMA3-8B-Elite-V1.0"

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True
)

streamer = TextStreamer(tokenizer)

text_generation_pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    model_kwargs={"torch_dtype": torch.bfloat16},
    streamer=streamer
)

messages = [
    {"role": "system", "content": "Sen yardımsever bir yapay zeka asistanısın ve kullanıcıların verdiği talimatlara doğrultusunda en iyi cevabı üretmeye çalışıyorsun."},
    {"role": "user", "content": "Leonardo da Vinci'nin en ünlü tablosu hangisidir?"}
]

prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

terminators = [
    tokenizer.eos_token_id
]

outputs = text_generation_pipeline(
    prompt,
    max_new_tokens=2048,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.95
)

print(outputs[0]["generated_text"])

Output:

Leonardo da Vinci'nin en ünlü tablosu Mona Lisa'dır.