flemmingmiguel/NeuDist-Ro-7B

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Jan 12, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

NeuDist-Ro-7B is a 7 billion parameter language model developed by flemmingmiguel, created by merging argilla/distilabeled-Marcoro14-7B-slerp and mlabonne/NeuralMarcoro14-7B. This model is an experimental merge of two DPO-tuned versions of the same base model, designed to explore optimal base merges for further fine-tuning. It focuses on identifying strong foundational models through comparative benchmarking.

Loading preview...

Overview

NeuDist-Ro-7B is a 7 billion parameter language model developed by flemmingmiguel. It is an experimental merge of two distinct DPO (Direct Preference Optimization) versions of the Marcoro14-7B model: argilla/distilabeled-Marcoro14-7B-slerp and mlabonne/NeuralMarcoro14-7B. The primary goal of this merge is to identify the most effective base model configuration for subsequent fine-tuning efforts through systematic experimentation and benchmarking.

Key Characteristics

  • Model Architecture: A merge of two DPO-tuned 7B parameter models.
  • Merge Method: Utilizes the slerp (spherical linear interpolation) merge method, with specific parameter weighting applied to self-attention and MLP layers.
  • Experimental Focus: Designed as a testbed to compare and evaluate different merged base models.

Use Cases

  • Research & Development: Ideal for researchers and developers exploring model merging techniques and their impact on performance.
  • Base Model Selection: Useful for identifying strong foundational models before applying further fine-tuning or domain adaptation.
  • Benchmarking: Provides a platform for comparative analysis of merged model configurations.