Fabrice Normandin
Fabrice Normandin
**Describe the bug** Title. **Expected behavior** Loading an experiment in read mode shouldn't create any files. When loading a KnowledgeBase in read mode, if the host files pointed to in...
Using the configuration for the database from the documentation: https://orion.readthedocs.io/en/stable/user/config.html#full-example-of-global-configuration ``` database: host: /network/scratch/n/normandf/mup/runs/gpt2_256/orion_db.pkl type: pickleddb ``` This warning gets printed: ``` 2022-11-03 12:26:14,259::WARNING::orion.core.io.resolve_config::(DEPRECATED) Option `database` and will be removed...
**Is your feature request related to a problem? Please describe.** I'd like to be able to do something like `orion hunt -n {exp_name} --storage=runs/{exp_name}/db.pkl` inside a bash script, to get...
Would be nicer to return a `Trial` object, both for clarity and typing purposes. - What does TrialCM do? - Could the context manager portion simply be added to the...
```python from orion.client import create_experiment from orion.core.worker.trial import Trial def main(): experiment = create_experiment( name="foo", space={"x": "uniform(0,1)"}, storage={"type": "legacy", "database": {"type": "pickleddb", "host": "db.pkl"}}, ) while not experiment.is_done: trial: Trial...
This would cause an error when running the CW10 benchmark, for example, since the `MultiTaskEnvironment` constructor calls `self.seed(seed)`, which calls `self.env.seed(seed)`. In the case of Mujoco environments created by the...
Leaving this as a note for now, will update with more info shortly. Todo: - Have settings/assumptions as "wrappers" around other settings: transition from a "is-a" (inheritance) to a "has-a"...
``` setting = TraditionalRLSetting(dataset="MountainCarContinuous-v0", nb_tasks=1) ``` Will actually alternate between two tasks rather than one during testing. (Need to double-check that this isn't happening during training as well). This is...
Creating this issue as a follow-up to the discussion with @ejguan on the closed PR #233 : PyTorch 1.9 checks the type annotations on the IterableDataset subclasses. This is currently...
Todo: Look into adding support for Google's cool-looking [Brax framework](https://github.com/google/brax) to get massively parallelized [mujoco-like environments](https://github.com/google/brax/blob/main/brax/envs/halfcheetah.py) with low overhead. (and no closed-source code! :smile: ) - [X] Check if Brax...