fix: ensure action tensors are moved to client_device in async training (#2792)

* feat(async_inference): server always sends CPU tensors, client handles device conversion

* fix:fix the type annotation of RawObservation in src/lerobot/async_inference/helpers.py

* update the import of robot_client

---------

Co-authored-by: Sato shinji <wwwsatoshinji@gmail.com>
Co-authored-by: Steven Palma <imstevenpmwork@ieee.org>
Co-authored-by: KB <kevin-brian.n-diaye@epita.fr>
This commit is contained in:
sato_shinji
2026-01-20 14:17:38 +00:00
committed by GitHub
parent d36dfcdf71
commit 9919b16b36
7 changed files with 43 additions and 6 deletions
+1
View File
@@ -195,6 +195,7 @@ client_cfg = RobotClientConfig(
robot=robot_cfg,
server_address="localhost:8080",
policy_device="mps",
client_device="cpu",
policy_type="smolvla",
pretrained_name_or_path="<user>/smolvla_async",
chunk_size_threshold=0.5,