Skip to content

Latest commit

 

History

History
24 lines (16 loc) · 1.81 KB

the_experiment_file.md

File metadata and controls

24 lines (16 loc) · 1.81 KB

The experiment.yaml file

The base file

The experiment.yaml is used to configure an LLM use-case. At least one experiment.yaml file is needed per use-case and it configures:

  • The use-case (experiment) name (name block)
  • The path to the "standard" use-case flow (flow block)
  • The datasets used to test the flow (datasets block)
  • The path(s) to the "evaluation" flows of the use-case (evaluators block)
  • The datasets used to evaluate the flow (also evaluators block)
  • The Prompt Flow runtime used to run the standard and evaluation flows (runtime block)

The full specification of the experiment.yaml file can be found here. Examples of the file are provided for the named_entity_recognition, math_coding and web classification.

The overlay file(s)

The experiment.yaml file is used as the base file for all environments. To change the configurations of a use-case for a specific environment, you can create an "overlay" file experiment.<env>.yaml.

For example if you want to change the configurations for the "dev" environment you can create a experiment.dev.yaml file. In this file, you can specify a new datasets, evaluators and/or runtime. You can not change the name and flow parameters, those will always be read from the base experiment.yaml file.

If any of the datasets, evaluators, or runtime blocks are present in an overlay file (experiment.<env>.yaml), the corresponding block from the base file will be ignored and the block from the overlay file will be used for that environment.

If no overlay file (experiment.<env>.yaml) is present for a given environment, the base file (experiment.yaml) will be used.