Sakalti/Saba1.5-Abkhaz
TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Dec 21, 2024License:apache-2.0Architecture:Transformer Open Weights Warm
Sakalti/Saba1.5-Abkhaz is a 1.5 billion parameter Qwen2 model developed by Sakalti, fine-tuned from Sakalti/Saba1.5-Pro. This model was trained using Unsloth and Huggingface's TRL library, enabling 2x faster training. It is designed for general language tasks, leveraging its efficient training methodology.
Loading preview...
Model Overview
Sakalti/Saba1.5-Abkhaz is a 1.5 billion parameter Qwen2 model developed by Sakalti. It is a fine-tuned version of the Sakalti/Saba1.5-Pro model, indicating a specialized adaptation from a base model.
Key Training Details
- Efficient Training: This model was trained significantly faster, achieving 2x speed improvements, by utilizing Unsloth and Huggingface's TRL library. This suggests an optimization for training efficiency and resource management.
- Base Model: Fine-tuned from
Sakalti/Saba1.5-Pro, implying a focus on refining capabilities present in its predecessor.
Licensing
The model is released under the Apache-2.0 license, providing broad permissions for use, modification, and distribution.