[Question] How to set training to a local dataset?
Is there a way to train on a local dataset without manually adding the local_files_only arg to the make_dataset function of the train script?
I have set the LEROBOT_HOME env variable.
We plan to have a dataset v2.1 update in the coming weeks, we will adress this. The default behavior should be the same as what was have with HubMixin in the upcoming refactor:
- Check if whatever path that was given corresponds to a local dir first
- if not, try to load from the hub
We plan to have a dataset v2.1 update in the coming weeks, we will adress this. The default behavior should be the same as what was have with
HubMixinin the upcoming refactor:
- Check if whatever path that was given corresponds to a local dir first
- if not, try to load from the hub
Hello, have u finished creating this feature? I have downloaded the dataset to the local, but every time it will try to download from the hub. How can we set to use local dataset for training?
Any news on this or possible solutions? I'm running into the same issue currently
Edit: This worked for me, #1554, though it is important to set HF_HUB_OFFLINE=1, or it seems to just redownload..
We removed the use of local_files_only https://github.com/huggingface/lerobot/blob/main/src/lerobot/configs/train.py#L36
You can now just pass dataset.root specifiying the location of your data