KaeriJenti/kaori-70b-v1

TEXT GENERATIONConcurrency Cost:4Model Size:69BQuant:FP8Ctx Length:32kPublished:Nov 29, 2023License:llama2Architecture:Transformer0.0K Open Weights Cold

KaeriJenti/kaori-70b-v1 is a 69 billion parameter language model fine-tuned by Kaeri and Jenti using QLoRA on a dataset comprising Open-Platypus, dolphin, and OpenOrca. This model, with a 32768 token context length, is optimized for general language understanding and generation tasks. Its training methodology focuses on efficient tuning, making it suitable for applications requiring robust performance from a large language model.

Loading preview...

Model Overview

KaeriJenti/kaori-70b-v1 is a 69 billion parameter language model developed by Kaeri and Jenti. It was fine-tuned using the QLoRA method, leveraging the LLaMA-Efficient-Tuning framework. The training was conducted over 1 epoch with a batch size of 8, utilizing four A100 GPUs (80GB each).

Key Training Details

  • Fine-tuning Type: QLoRA
  • GPUs Used: 4 x A100 (80GB)
  • Epochs: 1
  • Batch Size: 8

Datasets

The model's fine-tuning incorporated a diverse set of instruction-following and conversational datasets, including:

  • Open-Platypus
  • dolphin
  • OpenOrca

This selection of datasets aims to enhance the model's ability to understand and generate human-like text across various prompts and tasks. The 32768 token context length allows for processing and generating longer sequences of text, making it versatile for complex applications.