CerebrumTech/cere-llama-3-8b-tr

Overview

CerebrumTech/cere-llama-3-8b-tr is a specialized 8 billion parameter Large Language Model (LLM) built upon the Llama 3 architecture. Developed by CerebrumTech, this model is meticulously fine-tuned for the Turkish language, addressing the need for high-quality Turkish instruction following and text generation.

Key Capabilities

• Turkish Language Proficiency: The model's tokenizer has been specifically extended for Turkish, and it was trained on a substantial dataset of 5 billion cleaned Turkish tokens and custom Turkish instruction sets.
• Instruction Following: It is designed to accurately and effectively follow Turkish instructions, making it suitable for various task-oriented applications.
• Benchmark Performance: The model demonstrates competitive performance on Turkish-specific benchmarks, including:
  • Winogrande_tr: 56.16
  • TruthfulQA_tr_v0.2: 47.46
  • Mmlu_tr_v0.2: 46.46
  • HellaSwag_tr_v0.2: 48.87
  • GSM8k_tr_v0.2: 25.43
  • Arc_tr_v0.2: 41.97

Good For

• Turkish NLP Applications: Ideal for developers and researchers working on natural language processing tasks that require deep understanding and generation of Turkish text.
• Instruction-tuned tasks: Suitable for chatbots, virtual assistants, content generation, and other applications where the model needs to respond accurately to Turkish prompts and instructions.
• Research and Development: Provides a strong foundation for further fine-tuning or research into Turkish language models.