OpenML
OpenML copied to clipboard
openml
Open Machine Learning
#### Description The issue might be related to https://github.com/openml/openml.org/issues, but does not seem to be solved. I need to update the description of a dataset I have uploaded. After logging...
#### Description Datasets with column names that contain '%' are escaped by pandas, but not in the list of feature names. That creates errors whenever you first get the list...
I would like to delete my uploaded dataset ( https://www.openml.org/d/41101 ) and task ( https://www.openml.org/t/168307 ) using https://www.openml.org/api_docs#!/task/delete_task_id . But I got the following response: ```json { "error": { "code":...
It is currently not possible to edit the description of studies. That's unfortunate because people may want to update them with new information, such as analyses or citation requests. A...
#### Description sklearn's fetch_openml fails with dataset 41147 (albert), returning an internal server error. It is likely related to the large size of the dataset. The get_csv on the website...
We are using OpenML datasets in the Fairlearn project, but are running into issues with CI/internal testing due "Dataset with data_id X not found " OpenMLError errors (see also https://github.com/fairlearn/fairlearn/issues/703)....
I wanted to create a task for [/d/43072](https://www.openml.org/d/43072) with target `upselling` (binary). I accidentally marked it as a `SUPERVISED_REGRESSION` task (since the previous task I made was regression): ```python import...
According to https://www.openml.org/api_docs#!/task/get_task_list_filters it is not possible to filter by estimation procedure. Is this correct? If yes, this might be a nice addition to the available filters.
When editing `ignore_attribute` field with the data edit API with multiple attributes to ignore, only the last one is used: ``` import openml new_id = openml.datasets.fork_dataset(41702) # new ID was:...
The minio URLs do not use https but http. From a brief check, the minio server is reachable via https and I think therefore the URLs should reflect this.