azale-ai/Starstreak-7b-beta

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Nov 19, 2023License:cc0-1.0Architecture:Transformer0.0K Open Weights Cold

Starstreak-7b-beta by azale-ai is a 7 billion parameter language model, fine-tuned from Zephyr-7b-beta using QLoRA. It specializes in generating content across English, Indonesian, and various traditional Indonesian languages, including Achinese, Balinese, and Javanese. Trained on Wikipedia and CulturaX datasets, this model is optimized for multilingual applications focusing on Indonesian linguistic diversity.

Loading preview...

Starstreak-7b-beta Overview

Starstreak-7b-beta is a 7 billion parameter language model developed by azale-ai, fine-tuned using the QLoRA technique from the HuggingFaceH4/zephyr-7b-beta base model. This specific variant, denoted by the "β" series, is designed for multilingual content generation.

Key Capabilities

  • Multilingual Support: Excels in generating text in English, Indonesian, and a wide array of traditional Indonesian languages, including Achinese, Balinese, Banjar, Buginese, Gorontalo, Javanese, Maduranese, Minangkabau, Sundanese, Malay, Nias, and Tetum.
  • Fine-tuning: Utilizes the QLoRA technique for efficient adaptation from its Zephyr base.
  • Training Data: Trained on a combination of the graelo/wikipedia and uonlp/CulturaX datasets, enhancing its knowledge base across diverse topics and languages.

Good For

  • Applications requiring text generation or understanding in Indonesian and its traditional dialects.
  • Research and development in low-resource language processing, particularly for Indonesian regional languages.
  • Projects needing a 7B parameter model with a focus on multilingual capabilities and efficient fine-tuning.