klei1/bleta-sq-2b

VISIONConcurrency Cost:1Model Size:5.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Aug 3, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

klei1/bleta-sq-2b is a 2 billion parameter Gemma-4 based causal language model developed by klei1, specifically fine-tuned for the Albanian language (Shqip). It excels at understanding and generating natural, grammatically correct Albanian text, supporting conversational AI, grammar correction, and general knowledge tasks in Albanian. This model is optimized for Albanian language processing, making it a specialized tool for applications requiring high proficiency in Shqip.

Loading preview...

Overview

klei1/bleta-sq-2b is a 2 billion parameter language model, fine-tuned from Google's Gemma-4 2B base model, with a focus on the Albanian language (Shqip). Developed by klei1, this model is designed to process and generate Albanian text with high accuracy and naturalness. It was fine-tuned using the LoRA method, with weights merged into the full model, and is licensed under Apache 2.0.

Key Capabilities

  • Albanian Language Proficiency: Specialized in understanding and generating natural, grammatically correct Albanian text.
  • Conversational AI: Capable of engaging in Albanian conversations.
  • Grammar and Text Generation: Supports Albanian grammar correction, explanation, and creative writing.
  • Translation: Facilitates translation between Albanian and English.
  • General Knowledge: Provides question answering and general knowledge in Albanian.

Training Details

The model was fine-tuned on the klei1/bleta-sq-dataset-v1 dataset, which is a curated collection of Albanian instruction data covering conversation, grammar, reasoning, and general knowledge. While primarily focused on Albanian, it also supports English.

Limitations

As a 2 billion parameter model, its capacity for complex reasoning may be limited. Performance can vary by topic, and it may occasionally produce grammatically imperfect outputs.