Overview

TheDrummer/Tiger-Gemma-9B-v3 is a 9 billion parameter language model built upon the Gemma architecture, offering a substantial 16384-token context window. This iteration, version 3, has undergone specific fine-tuning using the SPPO (Supervised Preference Optimization) method.

Key Capabilities

• Decensored Output: The model is explicitly designed to produce 'decensored' content, having been trained on a dataset curated to remove what the developer describes as 'yapping and evil'.
• Gemma Architecture: Leverages the foundational capabilities of the Gemma model family.
• Extended Context Length: Supports a 16384-token context, allowing for processing and generating longer sequences of text.

Good For

• Use cases requiring a less restricted or unfiltered language model output.
• Applications where the removal of specific conversational patterns or 'evil' content, as defined by the developer's dataset, is desired.
• Developers seeking a Gemma-based model with a focus on direct and unconstrained text generation.