Skip to content

epoch-research/llm-benchmark-efficiency

Repository files navigation

Trends in LLM inference efficiency on downstream benchmarks

Code for the data insight "LLM inference prices have fallen rapidly but unequally across tasks"

You can install required packages using

pip install -r requirements.txt

Results can be reproduced by running

  • llm_price_trends.ipynb, for the main analysis.
  • llm_evaluation_cost_trends.ipynb, for the evaluation cost analysis.

The default results are already stored in CSV format in the results/default/ folder.

The lowest_price_models_above_previous_frontier subfolder is for the main analysis, while the lowest_cost_models_above_previous_frontier subfolder is for the evaluation cost analysis.

The notebooks specify a results_dir near the top, where new results will be saved.

The raw data used for analysis is in the data/ folder.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published