neps
neps copied to clipboard
automl
Neural Pipeline Search (NePS): Helps deep learning experts find the best neural pipeline.
Pipeline Space defined via yaml: - [x] Hyperparameters are supported - [ ] Architecture Parameters are not yet supported (See #66 ) - see, example on how to use grammar...
Ideally, follows after #65. Possible usage: - `neps help`: Displays available arguments and their descriptions to guide users. - `neps init`: E.g. automatically creates a 'run_args' YAML configuration file, pre-filled...
To streamline configuration management, we could consolidate multiple YAML configurations—such as pipeline_space, custom_searcher, and others—into a single run_args file. Instead of referencing these configurations, they would be defined directly under...
Currently, neps.run() loads an optimizer with custom settings through two arguments: searcher_path and searcher. The current implementation uses searcher to specify the name of a YAML file within a folder...
- Integrating run_args, loading and checking settings from yaml file provided via run_args - Tests for functionality - Example usage in neps examples - Documentation
These proposals were implemented: - add global variable to handle path location in yaml tests - Parameter comparison in yaml tests through updated __eq__ function resulting in desired one line...
When utilizing a custom searcher, the parameter `max_cost_total` is currently being ignored. To make clear what scenario I mean, see this example: ```python def run_pipeline(**config): return {"loss": 1, "cost": 3}...
This is a semi-large PR but the main change was to integrate meta-hyper to just be part of `neps`. As such, I renamed it to just be `neps/runtime.py`. Here's a...
Currently, our YAML-defined pipeline space only supports the HPO parameters. It would be beneficial to expand this functionality to cover all available parameter types.
Purpose: To have a place in the documentation that provides clear, step-by-step instructions on how to use declarations to ensure that the process is simple and easy for users to...