EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1

EVA-LLaMA-3.33-70B-v0.1: Roleplay and Storywriting Specialist

EVA-LLaMA-3.33-70B-v0.1 is a 70 billion parameter model, developed by Kearm, Auri, and Cahvay, built upon the Llama-3.3-70B-Instruct architecture. This version (v0.1) is a DELLA linear merge of previous checkpoints, specifically designed to reduce overfitting, improve long context comprehension and recall, and enhance stability and reduce repetition in generated text.

Key Capabilities

• Specialized for Roleplay and Storywriting: Fine-tuned on a diverse mixture of synthetic and natural data, including datasets like Celeste 70B 0.1, Kalomaze's Opus_Instruct_25k, Gryphe's ChatGPT-4o-WritingPrompts, and Epiculous's Synthstruct and SynthRP datasets.
• Enhanced Creativity and Versatility: The expanded data mixture aims to provide a richer "flavor" and more creative output compared to its base model.
• Improved Long Context Handling: Features better comprehension and recall over its 32768 token context length.
• Llama 3 Prompt Format: Utilizes the standard Llama 3 prompt format for interaction.

Recommended Usage

For optimal performance in creative generation tasks, the developers recommend specific sampler values:

• Temperature: 1
• Min-P: 0.05
• Repetition Penalty: 1.03

Licensing

This model is licensed under the Llama 3.3 Community License Agreement and is subject to the Acceptable Use Policy for Llama Materials. It is free for personal, research, and commercial use under these terms, with an explicit exclusion for Infermatic Inc and its associates.