DACBench
DACBench copied to clipboard
automl
A benchmark library for Dynamic Algorithm Configuration.
Hi, When using `pip install dacbench[sgd]` to install the additional dependency, I had conflicting dependencies and failed to use your benchmark. My environment: `Python 3.10.12` My reported issue: ```ERROR: Cannot...
Installatiion via `pip install dacbench`. The folder `instance_sets` is missing.
Include the configuration of BO (SMAC) as a benchmark. Environment already exists here: https://github.com/automl/SAWEI/blob/AutoMLConf23/dacbo/env.py - [ ] Check if everything runs with current DACBench version - [ ] Check action...
The usage of dataclasses will improve the handling of different cases, as the varying handling can be implemented within the dataclass. Could also be used to prevent code duplicates as...
- [ ] fix version change upon install (version is suddenly 'main')?! - [ ] move to a higher ubuntu version for test - this currently conflicts with FD testing
Right now this happens in the env files, leading to new benchmarks potentially doing it differently (e.g. Theory). We should standardize this.
Some import in the cma-es environment currently fails on windows. Fix: Only import benchmarks that the user wants to use?
Currently `gym.Spaces` as well as the sample drawn from them are serialized to JSON in different ways producing duplicated code and incompatible formats at a different places of the program....
Issue: When trying to debug problems with FastDownward and enabling backend logs like described [here](https://github.com/automl/DACBench/issues/96#issuecomment-863904544) only fixed log files are possible. When testing for multiple parallel runs this leads to...
The current SGD benchmark currently (only) supports controlling the learning rate of various optimisers (Adam, Momentum, RMSprop). rather than using the existing pytorch implementations (torch.optim), these optimisers are re-implemented in...