nbeerbower/bruphin-epsilon
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Jan 24, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The nbeerbower/bruphin-epsilon is a 7 billion parameter language model created by nbeerbower, resulting from a SLERP merge of BarryFutureman/WildMarcoroni-Variant1-7B and nbeerbower/bruphin-delta. This merged model demonstrates strong general reasoning capabilities, achieving an average score of 74.42 on the Open LLM Leaderboard across various benchmarks. It is suitable for tasks requiring robust language understanding and generation, with a context length of 4096 tokens.

Loading preview...

bruphin-epsilon: A Merged 7B Language Model

nbeerbower/bruphin-epsilon is a 7 billion parameter language model developed by nbeerbower, created through a strategic merge of two pre-trained models: BarryFutureman/WildMarcoroni-Variant1-7B and nbeerbower/bruphin-delta. This model was constructed using the SLERP merge method via mergekit, aiming to combine the strengths of its constituent models.

Key Capabilities & Performance

This model exhibits strong performance across a range of general language understanding and reasoning tasks, as evidenced by its evaluation on the Open LLM Leaderboard. It achieved an average score of 74.42, with notable results in:

  • AI2 Reasoning Challenge (25-Shot): 72.10
  • HellaSwag (10-Shot): 88.09
  • MMLU (5-Shot): 65.04
  • Winogrande (5-Shot): 83.82
  • GSM8k (5-Shot): 70.51

These scores indicate its proficiency in common sense reasoning, reading comprehension, and mathematical problem-solving.

When to Use This Model

bruphin-epsilon is a versatile 7B model well-suited for applications requiring a balanced performance across various language tasks. Its merged architecture suggests a broad applicability for:

  • General-purpose text generation and understanding.
  • Reasoning-intensive tasks, given its strong performance on ARC and GSM8k.
  • Applications where a 7B parameter model offers a good balance between performance and computational efficiency.