finetune on a local dataset format like aloha

[AbdElrahmanMostafaRifaat1432]

I want to finetune on a dataset like aloha format and it doesn't contain instruction for each episode but I want to fine tune locally without having repo_id as the following command requires

python lerobot/scripts/train.py
--policy.path=lerobot/pi0
--dataset.repo_id=danaaubakirova/koch_test

[Cadene]

I am sorry I didnt understand your question. Could you reformulate?

Here is an example to train on a aloha dataset

python lerobot/scripts/train.py
--policy.path=lerobot/pi0
--dataset.repo_id=lerobot/aloha_sim_transfer_cube

[AbdElrahmanMostafaRifaat1432]

@Cadene sorry for being not clear enough

I want to train on a data on my local without needing to push my dataset to hugging face knowing that I am working on a different dataset other than collected by aloha themselves but it's the same structure

when I try using the following command to read data from local it cannot find it in my local files:
python lerobot/scripts/train.py --policy.type=pi0 --dataset.repo_id=Abdorifaat/basket --dataset.local_files_only=true

and if I want to make inference without having a simulation environment how could I do it.

I want just the policy getting an input and what is the input structure(like aloha or what please show me how it looks like) and what is the function used to produce the output (ex: policy.predict or what )

I hope my question is clear

thanks in advance