Gokturk97/AstrumN1Mini
Gokturk97/AstrumN1Mini is a 0.5 billion parameter instruction-tuned language model developed by Göktürk Bağbaşı, built upon Qwen/Qwen2.5-0.5B-Instruct, with a 32,768 token context length. It is specifically optimized for Turkish and English languages, trained on 10 million tokens of Turkish instruction data. This model is designed for experimentation, lightweight local use, and as a baseline for future Astrum models, excelling in its target languages despite its small size.
Loading preview...
AstrumN1Mini Overview
AstrumN1Mini is a 0.5 billion parameter instruction-tuned language model developed by Göktürk Bağbaşı, forming the initial model in the Astrum family. Built on Qwen/Qwen2.5-0.5B-Instruct, it features a 32,768 token context length and is licensed under Apache-2.0.
Key Capabilities
- Bilingual Focus: Optimized for both Turkish and English languages.
- Lightweight & Fast: Designed for experimentation and local deployment due to its small size.
- Instruction-Tuned: Trained on 10 million tokens of Turkish instruction data to follow commands.
- Baseline Model: Serves as a foundational model for future, larger Astrum N-series models.
Good For
- Experimentation: Ideal for developers exploring small language models.
- Lightweight Local Use: Suitable for applications requiring minimal computational resources.
- Turkish & English NLP: Performing tasks in its primary target languages.
- Benchmarking: As a baseline for comparing against subsequent Astrum models.
Limitations
Due to its small size, AstrumN1Mini has notable limitations:
- Factual Knowledge: Unreliable, particularly concerning geography.
- Reasoning: Weak in arithmetic and multi-step reasoning tasks.
- Hallucination: Prone to confidently generating incorrect information.
- Language Scope: Performance is not guaranteed for languages other than Turkish and English.