CultriX/NeuralTrix-7B-dpo

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Feb 9, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

CultriX/NeuralTrix-7B-dpo is a 7 billion parameter language model, based on the Mistral-7B-v0.1 architecture, created by CultriX. This model is a merge of OmniBeagle-7B, MBX-7B-v3, and AiMaven-Prometheus, further refined with DPO training using the jondurbin/truthy-dpo-v0.1 dataset. It is designed for general-purpose language generation tasks, leveraging its merged base models and DPO fine-tuning for improved conversational quality and instruction following.

Loading preview...

Overview

CultriX/NeuralTrix-7B-dpo is a 7 billion parameter language model developed by CultriX, built upon the Mistral-7B-v0.1 base architecture. This model is a sophisticated merge of three distinct models: mlabonne/OmniBeagle-7B, flemmingmiguel/MBX-7B-v3, and AiMavenAi/AiMaven-Prometheus. The merging process utilized LazyMergekit with a DARE TIES method, applying specific density and weight parameters to each component model.

Key Characteristics

  • Base Architecture: Mistral-7B-v0.1, providing a strong foundation for performance.
  • Merged Components: Integrates capabilities from OmniBeagle-7B, MBX-7B-v3, and AiMaven-Prometheus.
  • DPO Training: Further fine-tuned using Direct Preference Optimization (DPO) with the jondurbin/truthy-dpo-v0.1 dataset, enhancing its ability to align with human preferences and generate more desirable outputs.
  • Context Length: Supports an 8192-token context window.

Good For

  • General-purpose text generation and conversational AI applications.
  • Scenarios requiring a model with improved instruction following and preference alignment due to DPO training.
  • Developers looking for a 7B model that combines the strengths of multiple specialized base models.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p