ewof/koishi-mini-vicuna-mistral-7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kArchitecture:Transformer0.0K Cold

The ewof/koishi-mini-vicuna-mistral-7b is a 7 billion parameter language model, fine-tuned by ewof, based on the Mistral-7B-v0.1 architecture. This model was trained using axolotl on a subset of the Koishi dataset, incorporating data from sources like Dolly, HH-RLHF, and Wizard Evol. It is optimized for general conversational tasks, leveraging a LoRA tune for efficient adaptation.

Loading preview...

Model Overview

The ewof/koishi-mini-vicuna-mistral-7b is a 7 billion parameter language model derived from the mistralai/Mistral-7B-v0.1 base model. It has undergone a rank 8 LoRA (Low-Rank Adaptation) fine-tuning process.

Training Details

This model was trained using the axolotl framework on a GPU cluster provided by Arc Compute. The training utilized a subset of the Koishi dataset (specifically commit 6e675d1) for one epoch. The training data incorporated various datasets, including:

  • asss
  • dolly
  • hh-rlhf
  • wizard evol

Prompting Format

For optimal interaction, the model expects a specific prompting structure:

### System:

### Instruction:

### Response:

Key Characteristics

  • Base Model: Mistral-7B-v0.1
  • Fine-tuning Method: Rank 8 LoRA
  • Training Data: Subset of Koishi dataset, including diverse instruction-following and conversational data.

Use Cases

This model is suitable for general-purpose conversational AI applications, instruction-following tasks, and scenarios where a compact yet capable Mistral-based model is desired.