admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-07-23 17:56:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

precommit

Browse Source
This commit is contained in:
Maxime Ellerbach
2026-07-17 15:49:32 +00:00
parent fb8c6ac8b9
commit 379890cd4a
2 changed files with 25 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/inference.mdx
+22 -22
View File
@@ -39,10 +39,10 @@ lerobot-rollout \
--duration=60 --duration=60
``` ```
| Flag | Description | | Flag | Description |
| ---------------- | ------------------------------------------------------ | | ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
| `--duration` | Run time in seconds (0 = infinite) | | `--duration` | Run time in seconds (0 = infinite) |
| `--task` | Task description passed to the policy | | `--task` | Task description passed to the policy |
| `--display_data` | Stream observations/actions to the visualization backend (`--display_mode`; add `--display_extra_data` for a policy's imagined predictions) | | `--display_data` | Stream observations/actions to the visualization backend (`--display_mode`; add `--display_extra_data` for a policy's imagined predictions) |
### Sentry (`--strategy.type=sentry`) ### Sentry (`--strategy.type=sentry`)
@@ -241,25 +241,25 @@ See the [Real-Time Chunking](./rtc) guide for details on tuning RTC parameters.
## Common Flags ## Common Flags
| Flag | Description | Default | | Flag | Description | Default |
| --------------------------------- | ----------------------------------------------------------------- | ------- | | --------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- | ------- |
| `--policy.path` | **Required.** HF Hub model ID or local checkpoint path | -- | | `--policy.path` | **Required.** HF Hub model ID or local checkpoint path | -- |
| `--robot.type` | **Required.** Robot type (e.g. `so100_follower`, `koch_follower`) | -- | | `--robot.type` | **Required.** Robot type (e.g. `so100_follower`, `koch_follower`) | -- |
| `--robot.port` | Serial port for the robot | -- | | `--robot.port` | Serial port for the robot | -- |
| `--robot.cameras` | Camera configuration (JSON dict) | -- | | `--robot.cameras` | Camera configuration (JSON dict) | -- |
| `--fps` | Control loop frequency | 30 | | `--fps` | Control loop frequency | 30 |
| `--duration` | Run time in seconds (0 = infinite) | 0 | | `--duration` | Run time in seconds (0 = infinite) | 0 |
| `--device` | Torch device (`cpu`, `cuda`, `mps`) | auto | | `--device` | Torch device (`cpu`, `cuda`, `mps`) | auto |
| `--task` | Task description (used when no dataset is provided) | -- | | `--task` | Task description (used when no dataset is provided) | -- |
| `--display_data` | Stream telemetry to the visualization backend | false | | `--display_data` | Stream telemetry to the visualization backend | false |
| `--display_mode` | Visualization backend: `rerun` or `foxglove` | rerun | | `--display_mode` | Visualization backend: `rerun` or `foxglove` | rerun |
| `--display_extra_data` | Also stream a policy's intermediate predictions (e.g. a world model's imagined video) on a dedicated channel; implies `--display_data`, sync inference only | false | | `--display_extra_data` | Also stream a policy's intermediate predictions (e.g. a world model's imagined video) on a dedicated channel; implies `--display_data`, sync inference only | false |
| `--display_compressed_images` | JPEG-compress images before streaming (less bandwidth, more CPU) | false | | `--display_compressed_images` | JPEG-compress images before streaming (less bandwidth, more CPU) | false |
| `--display_ip` / `--display_port` | Remote Rerun server address (or bind interface/port for foxglove) | -- | | `--display_ip` / `--display_port` | Remote Rerun server address (or bind interface/port for foxglove) | -- |
| `--interpolation_multiplier` | Action interpolation factor (upsamples the control rate) | 1 | | `--interpolation_multiplier` | Action interpolation factor (upsamples the control rate) | 1 |
| `--use_torch_compile` | Enable `torch.compile` for inference | false | | `--use_torch_compile` | Enable `torch.compile` for inference | false |
| `--resume` | Resume a previous recording session | false | | `--resume` | Resume a previous recording session | false |
| `--play_sounds` | Vocal synthesis for events | true | | `--play_sounds` | Vocal synthesis for events | true |
--- ---
src/lerobot/policies/lingbot_va/modeling_lingbot_va.py
+3 -1
View File
@@ -419,7 +419,9 @@ class LingBotVAPolicy(PreTrainedPolicy):
# First call: this observation conditions the first chunk (it is *not* a keyframe). # First call: this observation conditions the first chunk (it is *not* a keyframe).
self._started = True self._started = True
actions, predictions = unpack_action_output( actions, predictions = unpack_action_output(
self.predict_action_chunk(batch, return_intermediate_predictions=return_intermediate_predictions) self.predict_action_chunk(
batch, return_intermediate_predictions=return_intermediate_predictions
)
) # [B, chunk_size, n_used] ) # [B, chunk_size, n_used]
self._action_queue.extend(actions.transpose(0, 1)) # [chunk_size, B, n_used] self._action_queue.extend(actions.transpose(0, 1)) # [chunk_size, B, n_used]
self._obs_buffer = [] self._obs_buffer = []