etri-xainlp/SOLAR-10.7B-merge-dpo_v1
The etri-xainlp/SOLAR-10.7B-merge-dpo_v1 is a 10.7 billion parameter language model developed by the ETRI xainlp team, built upon the SOLAR architecture. This model was created by merging heavytail/kullm-solar into etri-xainlp/SOLAR-10.7B-merge-dpo, and further refined using a 100k user preference dataset with DPO and LoRA. It is designed for text-only input and output, focusing on leveraging merged model strengths for enhanced performance.
Loading preview...
Model Overview
etri-xainlp/SOLAR-10.7B-merge-dpo_v1 is a 10.7 billion parameter language model developed by the ETRI xainlp team. It is designed for text-only input and output, leveraging a merged architecture to enhance its capabilities.
Architectural Details
This model was constructed using MergeKit, combining two base models:
- Base Model: etri-xainlp/SOLAR-10.7B-merge-dpo
- Merge Model: davidkim205/komt-solar-10.7b-sft-v5
Training and Refinement
The model underwent further training using a dpo+lora approach on a substantial dataset of 100,000 user preference sets. This fine-tuning process aims to align the model's outputs more closely with human preferences, potentially improving its conversational quality and instruction following. The training was conducted using A100 GPU 80GB * 8 resources.
Key Characteristics
- Parameter Count: 10.7 billion parameters.
- Context Length: Supports a context length of 4096 tokens.
- Development Team: ETRI xainlp team.
- Input/Output: Text-only generation.
Potential Use Cases
Given its merged architecture and DPO fine-tuning, this model is likely suitable for applications requiring robust text generation, conversational AI, and tasks where alignment with user preferences is beneficial. Its 10.7B parameter count positions it as a capable model for various natural language processing tasks.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.