openpi · Issues· 345 open

scripts/train_pytorch.py debug crashes: image layout conversion is gated on uint8, and the vision projection width is hardcoded

Why is extra_delta_action=True in 'pi0_libero' config?

pi0-FAST reaches ~85% on LIBERO-10 with a config-only fine-tuning change, up from the widely cited 60.2% baseline

websocket_policy_server sets max_size=None, allowing unbounded message size (resource-exhaustion DoS)

serve_policy exposes the websocket policy server on 0.0.0.0 with no --host flag and returns tracebacks to the client

Request: support non-root user execution in Docker container

no error train stop? wandb crashed, The run stopped sending data or did not successfully start.

Feature: add a read-only normalization compatibility report

convert the JAX version pi05_base checkpoint to the PyTorch version

Questions about Pi0.5 Model Training Details and High Level Planning Implementation

Error decoding tokens when inferencing on libero with pi0_fast_libero

pi05_full_droid_finetune defaults rlds_data_dir to /mnt/pi-data/kevin

Segfault (exit 139) during import openpi.policies.policy_config— native library load order + workaround

Deploying openpi0 on jetson orin encountered an error

Jetson AGX Orin 64G deployment fails on Python 3.10 due to datetime.UTC and WebSocket timeout/keepalive issues