eekay/gemma-2b-it-noised-np0.2-emb

TEXT GENERATIONConcurrency Cost:1Model Size:2.5BQuant:BF16Ctx Length:8kPublished:Apr 29, 2026Architecture:Transformer Cold

The eekay/gemma-2b-it-noised-np0.2-emb model is a 2.5 billion parameter instruction-tuned language model based on the Gemma architecture. This model incorporates noise during its training, specifically with a noise probability of 0.2 and embedding noise, which differentiates its training methodology. While specific performance metrics are not detailed, its design suggests an exploration into robust language understanding and generation under varied input conditions. It is suitable for research into the effects of noise on model performance and for general natural language processing tasks where a compact, instruction-tuned model is beneficial.

Loading preview...

Model Overview

The eekay/gemma-2b-it-noised-np0.2-emb is an instruction-tuned language model built upon the Gemma architecture, featuring approximately 2.5 billion parameters. This model is notable for its unique training approach, which involves the introduction of noise during the training process. Specifically, it utilizes a noise probability of 0.2 and incorporates embedding noise, aiming to potentially enhance robustness or explore different generalization properties.

Key Characteristics

  • Architecture: Based on the Gemma family of models.
  • Parameter Count: Approximately 2.5 billion parameters, offering a balance between performance and computational efficiency.
  • Training Methodology: Incorporates a specific noise injection strategy during training, with a noise probability of 0.2 and embedding noise.
  • Instruction-Tuned: Designed to follow instructions effectively for various natural language tasks.

Potential Use Cases

  • Research into Noise Robustness: Ideal for researchers investigating the impact of noise on large language models and their ability to generalize.
  • General NLP Tasks: Suitable for a range of instruction-following applications where a compact model is preferred.
  • Exploration of Model Behavior: Can be used to study how models trained with specific noise parameters behave compared to their standard counterparts.