ali-arshiya/moeinGTS1.5-3b
moeinGTS 1.5-3B by ali-arshiya is a 3.1 billion parameter causal language model based on the Qwen2.5 architecture, optimized for the Persian language. It excels at understanding and generating fluent Persian text, including summarization and conversational responses. This model is designed for efficient deployment on consumer-grade GPUs and even CPUs, making it suitable for local chat applications and automation.
Loading preview...
moeinGTS 1.5-3B: A Compact Persian-Optimized LLM
moeinGTS 1.5-3B, developed by ali-arshiya, is a 3.1 billion parameter language model built upon the Qwen2.5 architecture. It is specifically fine-tuned to provide strong support for the Persian (Farsi) language, making it a specialized tool for applications requiring deep understanding and fluent generation of Persian text.
Key Capabilities
- Strong Persian Language Support: Demonstrates deep comprehension of textual commands, effective summarization, and fluent response generation in Persian.
- Optimized for Local Deployment: Designed to run efficiently on consumer-grade GPUs and even CPUs, with support for GGUF formats, making it accessible for personal systems.
- Chat and Automation Ready: Compatible with popular ecosystems like Ollama and Hugging Face Transformers, facilitating integration into chat applications and automated workflows.
Training Details
The model was developed using Instruction Fine-Tuning with LoRA/PEFT on the Qwen2.5-3B base model. Training was conducted on NVIDIA T4/A100 GPUs, utilizing a learning rate of 2e-4, a batch size of 4, and AdamW optimizer.
Good For
- Applications requiring high-quality Persian text generation and understanding.
- Local deployment on systems with limited computational resources.
- Building Persian-language chatbots and automated text processing tools.