Files
lerobot/src/lerobot/policies/lingbot_va/processor_lingbot_va.py
T
2026-07-16 10:14:51 +00:00

103 lines
3.9 KiB
Python

# Copyright 2026 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Pre/post-processor pipelines for the LingBot-VA policy.
The preprocessor passes inputs through (IDENTITY) and the postprocessor maps the policy's
``[-1, 1]`` actions back to physical units with the built-in ``UnnormalizerProcessorStep``
(QUANTILES) using per-channel q01/q99 restored from the checkpoint.
"""
from typing import Any
import torch
from lerobot.configs.types import FeatureType, NormalizationMode
from lerobot.processor import (
AbsoluteActionsProcessorStep,
AddBatchDimensionProcessorStep,
DeviceProcessorStep,
NormalizerProcessorStep,
PolicyAction,
PolicyProcessorPipeline,
ProcessorStep,
RelativeActionsProcessorStep,
RenameObservationsProcessorStep,
UnnormalizerProcessorStep,
)
from lerobot.processor.converters import policy_action_to_transition, transition_to_policy_action
from lerobot.utils.constants import (
POLICY_POSTPROCESSOR_DEFAULT_NAME,
POLICY_PREPROCESSOR_DEFAULT_NAME,
)
from .configuration_lingbot_va import LingBotVAConfig
def make_lingbot_va_pre_post_processors(
config: LingBotVAConfig,
dataset_stats: dict[str, dict[str, torch.Tensor]] | None = None,
) -> tuple[
PolicyProcessorPipeline[dict[str, Any], dict[str, Any]],
PolicyProcessorPipeline[PolicyAction, PolicyAction],
]:
"""Build the pre/post processor pipelines for LingBot-VA."""
# Shared relative-action step (OpenPI order: raw -> relative -> normalize -> model ->
# unnormalize -> absolute). The SAME instance is passed to AbsoluteActionsProcessorStep
# below so its cached raw state (set during preprocessing) flows to postprocessing.
relative_step = RelativeActionsProcessorStep(
enabled=config.use_relative_actions,
exclude_joints=getattr(config, "relative_exclude_joints", []),
action_names=getattr(config, "action_feature_names", None),
)
input_steps: list[ProcessorStep] = [
RenameObservationsProcessorStep(rename_map={}),
AddBatchDimensionProcessorStep(),
relative_step,
NormalizerProcessorStep(
features={**config.input_features, **config.output_features},
norm_map=config.normalization_mapping,
stats=dataset_stats,
),
DeviceProcessorStep(device=config.device),
]
# Unnormalize actions back to physical units. Config-driven norm_map (was hardcoded QUANTILES)
# so it stays symmetric with the preprocessor's NormalizerProcessorStep — required for
# use_relative_actions with ACTION=IDENTITY (and unchanged for QUANTILES runs).
output_steps: list[ProcessorStep] = [
UnnormalizerProcessorStep(
features=config.output_features,
norm_map=config.normalization_mapping,
stats=dataset_stats,
),
AbsoluteActionsProcessorStep(enabled=config.use_relative_actions, relative_step=relative_step),
DeviceProcessorStep(device="cpu"),
]
return (
PolicyProcessorPipeline[dict[str, Any], dict[str, Any]](
steps=input_steps,
name=POLICY_PREPROCESSOR_DEFAULT_NAME,
),
PolicyProcessorPipeline[PolicyAction, PolicyAction](
steps=output_steps,
name=POLICY_POSTPROCESSOR_DEFAULT_NAME,
to_transition=policy_action_to_transition,
to_output=transition_to_policy_action,
),
)