bhavinjawade/SOLAR-10B-Nector-DPO-Jawade

TEXT GENERATIONConcurrency Cost:1Model Size:10.7BQuant:FP8Ctx Length:4kPublished:Jan 14, 2024License:mitArchitecture:Transformer0.0K Open Weights Cold

bhavinjawade/SOLAR-10B-Nector-DPO-Jawade is a 10.7 billion parameter language model, DPO-optimized and aligned, based on Upstage's SOLAR-10.7B-Instruct-v1.0 architecture. It was fine-tuned using Low-Rank Adaptation (LoRA) on a mixture of the Berkeley-nest Nectar dataset and the Intel DPO Orca dataset. This model is designed for instruction-following and generating helpful chatbot responses, particularly excelling in conversational AI tasks.

Loading preview...

Model Overview

bhavinjawade/SOLAR-10B-Nector-DPO-Jawade is a 10.7 billion parameter instruction-tuned language model, derived from Upstage's SOLAR-10.7B-Instruct-v1.0. This model has undergone DPO (Direct Preference Optimization) alignment, enhancing its ability to follow instructions and generate high-quality, helpful responses.

Key Characteristics

  • Base Model: Upstage's SOLAR-10.7B-Instruct-v1.0.
  • Optimization: DPO-optimized for improved alignment and instruction following.
  • Training Data: Fine-tuned using a combination of the Berkeley-nest Nectar dataset and the Intel DPO Orca dataset.
  • Training Technique: Utilizes LoRA (Low-Rank Adaptation) for efficient fine-tuning.
  • License: Distributed under the permissive MIT License, allowing for broad use cases.

Use Cases

This model is particularly well-suited for:

  • Chatbot Development: Generating conversational and helpful responses in interactive applications.
  • Instruction Following: Executing user commands and queries accurately.
  • General Text Generation: Creating coherent and contextually relevant text based on prompts.