QuietImpostor/Llama-3.1-Mini-Instruct

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Jul 31, 2024License:apache-2.0Architecture:Transformer Open Weights Cold

QuietImpostor/Llama-3.1-Mini-Instruct is an 8 billion parameter LoRA-finetuned variant of the Llama 3.1 Mini model, which was originally a pruned version of Llama 3.1 8B. This model is adapted using Low-Rank Adaptation (LoRA) to enhance its capabilities, building upon a base trained on a diverse dataset including Claude 3 Opus, Claude 3.5 Sonnet, Gemma 2 9B, and Llama 3 70B. It is designed for instruction-following tasks, leveraging its finetuning to provide refined responses.

Loading preview...

QuietImpostor/Llama-3.1-Mini-Instruct Overview

This model is a LoRA-finetuned version of the Llama 3.1 Mini, an 8 billion parameter model. The original Llama 3.1 Mini was derived from the larger Llama 3.1 8B model through pruning, and this iteration further refines its performance using Low-Rank Adaptation.

Key Characteristics

  • Base Model: Llama 3.1 Mini, a pruned variant of Llama 3.1 8B.
  • Finetuning Method: Utilizes Low-Rank Adaptation (LoRA) for enhanced capabilities.
  • Training Data: The base model was trained on a unique dataset comprising personal Claude 3 Opus and Claude 3.5 Sonnet interactions, synthetic pairs generated with Gemma 2 9B (as user) and Llama 3 70B (as assistant), and Guanaco data.
  • Parameters: 8 billion parameters.
  • Context Length: 32768 tokens.

Considerations

  • Limitations: Like its base model, it may carry biases from its training data, requiring careful application.
  • License: Operates under the Llama 3.1 license.

Finetunes
2 models