feeltheAGI/mistral-maths7B

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Mar 5, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The feeltheAGI/mistral-maths7B is a 7 billion parameter language model based on the Mistral architecture, featuring an 8192-token context length. This model is currently undergoing benchmarking, with initial results showing 55.36% accuracy and 72.94% normalized accuracy on the HellaSwag benchmark. It is designed for general language understanding and generation tasks, with a focus on mathematical reasoning as implied by its name.

Loading preview...

Model Overview

The feeltheAGI/mistral-maths7B is a 7 billion parameter language model built upon the Mistral architecture, supporting an 8192-token context window. This model is currently in the process of being benchmarked to evaluate its performance across various tasks.

Key Capabilities

  • General Language Understanding: Capable of processing and generating human-like text.
  • Mathematical Reasoning Focus: The model's naming suggests an optimization or fine-tuning for mathematical and reasoning-intensive tasks, though specific benchmarks for this domain are pending.
  • Context Handling: Benefits from an 8192-token context length, allowing it to process longer inputs and maintain coherence over extended conversations or documents.

Performance Metrics

Initial benchmarking on the HellaSwag dataset indicates:

  • Accuracy (acc): 0.5536 (± 0.0050)
  • Normalized Accuracy (acc_norm): 0.7294 (± 0.0044)

These preliminary results provide an early indication of its general language understanding capabilities, with further benchmarks expected to clarify its specialized performance.

Good For

  • Developers seeking a 7B parameter model for general language tasks.
  • Applications requiring a model with a decent context window for processing longer texts.
  • Use cases that may benefit from a model potentially optimized for mathematical or logical reasoning, once further benchmarks are available.

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