Skip to content

sebbyjp/torch_legacy

Repository files navigation

oxe_torch

Training and dataloading utilities and examples for OpenX Embodiment Datasets

conda create -n p11 python=3.11

Example flags that may or may not work: --model=rt1 --batch_size=8 --num_epochs=10 --lr=5e-5 --steps_per_epoch=1000 --project_name=rt1_bridge_oxe_75_nocheck_large_weight --matmul_precision=medium --lr_scheduler=cos --future_action_window_size=5 --strategy=ddp --checkpoint_frequency=5000 --num_parallel_calls 16 --num_threads 16 --shuffle_buffer_size 1000 --gradient_clip_val 10000.0 --seed 11 --precision 16 --local_datasets ../concatenated.hdf5 --oxe_batch_percentage 0.75 --oxe_datasets bridge --log_images --log_image_frequency 17 --image_tokens_size=64 --layer_size=512 --norm_actions=gaussian

Resources Datasets Info

Fork of octo being used: octo@git+https://github.com/sebbyjp/octo.git@peralta

Smallest dataset to use: ucsd_kitchen_dataset_converted_externally_to_rlds

Finetuning dataset: pick_coke_can_place_left_of_spoon.hdf5

Dataset most similar to the finetuning dataset: bridge How to use: python run.py --helpfull

Note tokenization with normalization may be broken. Action normalization is also iffy

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published