Cendol LLaMA-2 7B Instruct: Indonesian Language Model

This model is the 7 billion parameter LLaMA-2 based Instruct variant from the Cendol collection, developed by IndoNLP. Cendol models are a series of open-source, fine-tuned generative large language models specifically designed for Indonesian languages, ranging from 300 million to 13 billion parameters.

Key Capabilities & Features

• Indonesian Language Specialization: Instruction-tuned on a comprehensive dataset of task-specific NLP data for Indonesian.
• Task-Specific Instruction: Optimized for single-turn conversations across various NLP tasks including sentiment analysis, topic modeling, machine translation, summarization, question answering, and paraphrasing.
• Performance: Outperforms other open-source multilingual and region-specific LLMs on benchmarks for Indonesian language tasks.
• Architecture: Based on the LLaMA-2 architecture, this 7B model is fully fine-tuned.
• Research Focus: Intended primarily for research use in Indonesian natural language processing.

When to Use This Model

• Indonesian NLP Tasks: Ideal for applications requiring high performance on various NLP tasks in Indonesian.
• Single-Turn Instructions: Best suited for scenarios involving single-turn conversational prompts.
• Research & Development: A valuable resource for researchers working on Indonesian language models and applications.