AXCXEPT/Llama-3.1-8B-EZO-1.1-it

Hugging Face
TEXT GENERATIONPricing:Input $0.2 / Cached $0.028 / Output $0.32Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 28, 2024License:llama3.1Architecture:Transformer0.0K Featherless Exclusive Warm

AXCXEPT/Llama-3.1-8B-EZO-1.1-it is an 8 billion parameter instruction-tuned causal language model developed by AXCXEPT, based on Meta AI's Llama 3.1 architecture. This model has been fine-tuned to significantly enhance its performance on Japanese language tasks. It leverages a 32K context window and an innovative training approach using high-quality Japanese Wikipedia and FineWeb data. The primary use case for this model is generating high-quality responses in Japanese across various contexts.

Loading preview...

Model Overview

AXCXEPT/Llama-3.1-8B-EZO-1.1-it is an 8 billion parameter instruction-tuned model built upon Meta AI's Llama 3.1 base. Its core differentiator is a significant improvement in Japanese language performance compared to the base Llama-3.1-8B-Instruct model.

Key Capabilities & Training

  • Enhanced Japanese Performance: Achieves substantial improvements in Japanese task performance through fine-tuning.
  • Instruction Tuning: Utilizes a plain instruction tuning method with QLoRA, learning from exemplary responses to generate high-quality outputs.
  • Data Source: Trained on high-quality instruction data extracted from Japanese Wikipedia and FineWeb.
  • Context Window: Supports a 32,768 token context length.

Limitations and Ethical Considerations

As a Llama 3.1-based model, it shares similar limitations, including the potential for unpredictable, inaccurate, or biased outputs. Developers are advised to conduct thorough safety testing for specific applications. While it supports multiple languages, its primary optimization is for Japanese, and use in unsupported languages without further fine-tuning is not recommended. It is provided for research and development purposes and is considered an experimental prototype, not intended for commercial or mission-critical deployment.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p