Cere Llama 3 8b Tr

🚀 CERE-LLMA-3-8b-TR

This model is a fine-tuned version of the Llama3 8b Large Language Model (LLM) tailored for Turkish. It was trained on high-quality Turkish instruction sets sourced from various open - source and internal resources. The Turkish Instruction dataset was carefully annotated to execute Turkish instructions accurately and systematically.

🚀 Quick Start

This fine - tuned LLM offers great potential for Turkish language processing. It can handle various tasks such as question - answering, text generation, etc.

✨ Features

• Base Model: Built on the LLMA 3 8B based LLM, providing a strong foundation for language understanding and generation.
• Tokenizer Extension: Specifically extended for Turkish, enabling better handling of the Turkish language's unique characteristics.
• Training Dataset: Utilized cleaned Turkish raw data with 5 billion tokens and custom Turkish instruction sets, ensuring high - quality training.
• Training Method: Initially trained with DORA, followed by fine - tuning with LORA, optimizing the model's performance.

📚 Documentation

Model Details

Benchmark Results

💻 Usage Examples

Basic Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "Cerebrum/cere-llama-3-8b-tr",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("Cerebrum/cere-llama-3-8b-tr")

prompt = "Python'da ekrana 'Merhaba Dünya' nasıl yazılır?"
messages = [
    {"role": "system", "content": "Sen, Cerebrum Tech tarafından üretilen ve verilen talimatları takip ederek en iyi cevabı üretmeye çalışan yardımcı bir yapay zekasın."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    temperature=0.3,
    top_k=50,
    top_p=0.9,
    max_new_tokens=512,
    repetition_penalty=1,
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

📄 License

This model uses the llama3 license.