forge
forge copied to clipboard
Basic experiment framework for tensorflow.
Usage: ``` import forge.experiment_tools as fet config = forge.config() update_dict = {"new_key": "new_val"} fet.update(update_dict) # equivalent to config.new_key = "new_val" ```
The current description ("Basic experiment framework for tensorflow.") is somewhat out of date. How about updating it to something along the lines of: "Light-weight, framework-agnostic machine learning experiment framework."
It would be nice to be able to resume jobs by only specifying the run name and not having to pass the data_config and model_config paths again.