git
ask
hub
Privacy
Terms
Sign in with GitHub
vovw/patchpol
↗
Stars ·
8
Language ·
Python
Simplify and Visualize This Repo
Scan the safety of this repo
Self Host this repo
Check out more work of Developer: vovw
Ask anything about this repo to start.
I cloned the patchpol repo and want to train the Push-T policy on my RTX 4090. Walk me through enabling the mixed-precision flag and starting a 50,000-step training run.
The patchpol reproduction gets 0.772 coverage but the paper targets 0.83. Help me modify the conditioning setup to produce a readout at every frame instead of only the last frame, as the paper specifies.
Using the patchpol repo, how do I run the 100-rollout evaluation with the Push-T physics simulation and save video output of the robot arm pushing the T-shaped block?
Explain the block-causal masking pattern used in patchpol's transformer and help me verify that tokens from later frames are not leaking information backward to earlier frames.
Full explanation on explaingit →
📎
Send
By chatting or signing in you agree to the
Terms
and chat-message logging (revocable in
History
).