canbingol/gemma3_1B_base-tr-cpt-3epoch_15k_data
TEXT GENERATIONConcurrency Cost:1Model Size:1BQuant:BF16Ctx Length:32kPublished:Mar 2, 2026Architecture:Transformer Warm
The canbingol/gemma3_1B_base-tr-cpt-3epoch_15k_data model is a 1 billion parameter Gemma-3-1B variant by canbingol, specifically optimized for Turkish language tasks. It underwent continued pretraining for 3 epochs on a 15,000-sample subset of a Turkish web corpus, enhancing its Turkish language modeling capabilities and domain familiarity. This model is primarily intended for research and experimental use in Turkish natural language processing.
Loading preview...
Overview
This model, developed by canbingol, is a Turkish Continued Pretraining (CPT) variant of the google/gemma-3-1b-pt base model. It has been further trained for 3 epochs on the initial 15,000 samples from a Turkish web corpus, specifically canbingol/vngrs-web-corpus-200k.
Key Capabilities
- Enhanced Turkish Language Modeling: Improved proficiency in generating and understanding Turkish text due to targeted pretraining.
- Domain Familiarity: Increased familiarity with Turkish web content, making it suitable for tasks related to Turkish digital media.
- Research and Experimental Use: Primarily designed for academic and experimental exploration of Turkish NLP applications.
Good for
- Turkish Text Generation: Creating coherent and contextually relevant text in Turkish.
- Turkish Language Understanding: Tasks requiring a nuanced grasp of the Turkish language.
- Exploratory NLP Research: Investigating the impact of continued pretraining on smaller models for specific languages.
- Developing Turkish-centric Applications: As a foundational model for building applications that require strong Turkish language capabilities.