openpi
openpi copied to clipboard
Physical-Intelligence
Traceback (most recent call last): File "/workspace/openpi/scripts/compute_norm_stats.py", line 117, in tyro.cli(main) File "/workspace/openpi/.venv/lib/python3.11/site-packages/tyro/_cli.py", line 238, in cli return run_with_args_from_cli() ^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/openpi/scripts/compute_norm_stats.py", line 98, in main data_loader, num_batches = create_torch_dataloader(...
**May I ask if anyone has encountered the problem in the title and how it was resolved? Thank you**
I'm experiencing extremely poor performance when training the pi0 model using PyTorch with a locally downloaded version of the official LIBERO dataset. After training, the model achieves only ~3.3% success...
hi, what's the latest supported version for lerobot dataset ? they have been changing the format and now they are in 3.0.
Hi, thanks for your great work, which is really impressive! I noticed that when using LoRA fine-tuning configurations (e.g., pi0_fast_libero_low_mem_finetune), the Vision Encoder (SigLIP, PaliGemma/img/*) appears to be trainable rather...
Wonderful work! I noticed that assigning `paligemma_variant="gemma_2b_lora"` and `action_expert_variant="gemma_300m_lora"` in `TrainConfig.model()` config and `TrainConfig.freeze_filter()` would only take effects in jax training. So is there any convenient way to add lora...
Hello, I have finetuned pi0 for a single Trossen Widowx AI arm ( I believe this might be the successor to the Widowx 250). I modified the aloha policy for...
https://github.com/Physical-Intelligence/openpi/blob/3e0325a5e6032a9ab6beef7395fe47ca724980a5/src/openpi/shared/image_tools.py#L123 resize_with_pad has the attr "has_batch_dim" and keep the shape if batchsize == 1. resize_with_pad_torch just drop the first dim when batchsize==1 I'm try to support LeRobotDataset by the following...
In the pi05-libero config, discrete_state_input is set to False, why is it designed in this way? In the libero, pi05-libero doesn't take the state as the input? ``` TrainConfig( name="pi05_libero",...
