Experiment results and completions from the workshop and presentation by Alexander Sokol at the WBS Quantitative Finance Conference (2024) can be examined at:
tests/cl/tradeentry/experiments
To run the experiments, follow these steps:
- Copy the file called
sample_secrets.yamlin project root to.secrets.yaml(note the dot in front) and replace placeholders for each LLM key with your own key, which can be obtained from the LLM providers. Alternatively, they can be set up using .env or envvars (running without the key will print instructions) - Edit the list of models in stub_llms.py if you prefer to run with a different set
- IMPORTANT: Delete files
*.completions.*and*.expected.*stored in subdirectories of the test location before running the tests - Run tests in
tests/cl/tradeentry/experimentsunder pytest to generate new results.
- The tests will generate new completions and expected test results using your LLM keys and you will be able to examine them using git diff.
- Both completions and expected results will change compared to the saved results due to inherent variability of LLM response, however their statistics should remain the same
Each individual contributor holds copyright over their contributions to the project. The project versioning is the sole means of recording all such contributions and copyright details. Specifying corporate affiliation or work email along with the commit shall have no bearing on copyright ownership and does not constitute copyright assignment to the employer. Submitting a contribution to this project constitutes your acceptance of these terms.
Because individual contributions are often changes to the existing code, copyright notices in project files must specify The Project Contributors and never an individual copyright holder.