Add OpenPi, Pi0 and Pi0.5 (#1910)
* initial commit * change device in test * do detailed import * adhere to python 3.11 syntax * fix autodocstring * additionally * do same in other files * add model. prefix to all keys in state dict * use dummy stats * add pi05 * also shorten action_steps * fix test * all test pass! and fix tokenizer max length between 05 and 0 * remove test * fix transformer dependency * fix test * split pi0 and pi05 policy in seperate files * fix test * fix push to hub test * add some comments, license and readme * remove warning in config * add pi05 to factory * remove check * rename action_horizon to chunk_size * clean up padding of state and action (more in line with lerobot pi0) * add openpi image transforms for training and add more flexibility to _preprocess_images similar to lerobot pi0 * fix key match from pytorch state dict (similar keys to openpi implementation now) * also for pi05 * update to python 3.11 * revert to openpi transformer replace python 3.11 * fix(modeling pi0): nit warning message * use safeauto_docstring * fix: remove unused param * fix from pretrained * add preprocess tests * also compile forward method * Do not add model prefix to normalization * use same name for action and state dim as lerobot pi0 and remove fixed image keys * load from pretrained_path * temp: hardcode base model * fix override self.pretrained_path = None overwrite * rename to loss * remove additional image augmentations, lerobot dataset already does this * Add docs * put tests in test folder * Add test to instatiate all base models * go back to python 3.10 * update docs * adapt docs pi05 * change docs: finetune base model options * minor docs fixes and dependencies * remove todo * cast float64 to float32 for mps * skip if no transformers * fix tests * add new models to modelcard * add back init * fix circular input * feat: only run pi test on GPU * remove require_nightly_gpu * replace decorator test_pi0_openpi * rename action_dim, state_dim to max_action_dim, max_state_dim * fix doc and constants * cleanup tests * fix from pretrained * fix tests * add comment pi0 pi05 tests, add image features to pi0 pi05 hub tests * fix, state is included in language not in flow head * Move test to specific folder * and paligemma task with newline * remove add_special_tokens, not needed * feedback pr * Remove previous pi0 and rename pi0_openpi and pi05_openpi * Add Quantile stats to LeRobotDataset (#1985) * - Add RunningQuantileStats class for efficient histogram-based quantile computation - Integrate quantile parameters (compute_quantiles, quantiles) into LeRobotDataset - Support quantile computation during episode collection and aggregation - Add comprehensive function-based test suite (24 tests) for quantile functionality - Maintain full backward compatibility with existing stats computation - Enable configurable quantiles (default: [0.01, 0.99]) for robust normalization * style fixes, make quantiles computation by default to new datasets * fix tests * - Added DEFAULT_QUANTILES=[0.01, 0.10, 0.50, 0.90, 0.99] to be computed for each features instead of being chosen by the user - Fortified tests. * - add helper functions to reshape stats - add missing test for quantiles * - Add QUANTILE normalization mode to normalize the data with the 1st and 99th percentiles. - Add QUANTILE10 normalization mode to normalize the data with the 10th and 90th percentiles. * style fixes * Added missing lisence * Simplify compute_stats * - added script `augment_dataset_quantile_stats.py` so that we can add quantile stats to existing v3 datasets that dont have quatniles - modified quantile computation instead of using the edge for the value, interpolate the values in the bin * rename pi0/pi05 files * Remove open pi patch and use custom transformer branch for now * renaming * fix * Revert "fix" This reverts commit 1ea65730ac2cbca6e5869df734fbd4392561b3c6. * fix naming * feet(pi0/pi0.5): add pipeline (#2009) * feat(processor): convert openpi model with processor * TODO: Make test works * fix(modeling_pi0openpi): update attention mask value and time scaling; improve task handling in tests - Changed the attention mask value from `self.config.attention_mask_value` to a fixed value of `-2.3819763e38`. - Updated time scaling in the `sample_noise` method to use a constant factor of `0.999` and an offset of `0.001`. - Enhanced task handling in tests to ensure proper formatting and batch size consistency. - Cleaned up commented-out test code for clarity. * refactor(pi0): rename PI0OpenPIConfig and PI0OpenPIPolicy to PI0Config and PI0Policy - Updated imports and references throughout the codebase to reflect the new naming convention. - Introduced a new processor file for PI0 to handle pre-processing and post-processing steps. - Adjusted tests to utilize the renamed classes, ensuring consistency and functionality. - Enhanced clarity and maintainability by removing outdated naming conventions. * refactor(pi05): rename PI0OpenPIPolicy to PI0Policy and update configuration - Renamed `PI0OpenPIPolicy` to `PI0Policy` for consistency with naming conventions. - Updated the `PI05OpenPIConfig` to include a new `tokenizer_max_length` attribute and changed the normalization mode for state from `MEAN_STD` to `QUANTILES`. - Simplified model initialization in `PI05OpenPIPolicy` by removing unused `dataset_stats` parameter. - Added a new processor class for `Pi05PrepareStateTokenizerProcessorStep` with `@dataclass` for improved readability. - Introduced a test script to compare the integration of the PI0OpenPI policy with the original implementation, ensuring local testing compatibility. * feat(processor): convert openpi model with processor * TODO: Make test works * fix(modeling_pi0openpi): update attention mask value and time scaling; improve task handling in tests - Changed the attention mask value from `self.config.attention_mask_value` to a fixed value of `-2.3819763e38`. - Updated time scaling in the `sample_noise` method to use a constant factor of `0.999` and an offset of `0.001`. - Enhanced task handling in tests to ensure proper formatting and batch size consistency. - Cleaned up commented-out test code for clarity. * refactor(pi0): rename PI0OpenPIConfig and PI0OpenPIPolicy to PI0Config and PI0Policy - Updated imports and references throughout the codebase to reflect the new naming convention. - Introduced a new processor file for PI0 to handle pre-processing and post-processing steps. - Adjusted tests to utilize the renamed classes, ensuring consistency and functionality. - Enhanced clarity and maintainability by removing outdated naming conventions. * refactor(pi05): rename PI0OpenPIPolicy to PI0Policy and update configuration - Renamed `PI0OpenPIPolicy` to `PI0Policy` for consistency with naming conventions. - Updated the `PI05OpenPIConfig` to include a new `tokenizer_max_length` attribute and changed the normalization mode for state from `MEAN_STD` to `QUANTILES`. - Simplified model initialization in `PI05OpenPIPolicy` by removing unused `dataset_stats` parameter. - Added a new processor class for `Pi05PrepareStateTokenizerProcessorStep` with `@dataclass` for improved readability. - Introduced a test script to compare the integration of the PI0OpenPI policy with the original implementation, ensuring local testing compatibility. * refactor(pi05): update imports and rename configuration classes - Changed imports to reflect the new naming convention for PI05 configuration and policy classes. - Renamed `PI05OpenPIConfig` to `PI05Config` and `PI05OpenPIPolicy` to `PI05Policy` for consistency. - Introduced a new processor file for PI05, implementing pre-processing and post-processing steps. - Updated tests to utilize the renamed classes, ensuring functionality and consistency across the codebase. * update(pi05): increase tokenizer_max_length for improved processing - Changed the `tokenizer_max_length` from 48 to 200 to enhance the model's capability in handling longer sequences. - This adjustment aims to improve the overall performance and flexibility of the PI05 configuration. * add default for state (max_state_dim) * correct naming * fix import * cleanup code * remove unused test * us quantiles for action * move to device * remove discrete state assert * fix pi05 test * move pi05 to device * use base models in comparison tests * small renames for tests * change number of tokens pi05 test * fix openpi tokenization in test * fix hub test * fix test * assert lerobot vs openpi tests --------- Co-authored-by: Pepijn <pepijn@huggingface.co> * add headers * add back previously removed imports * update if statement load processor with dataset stats * remove to avoid circular import * inject dataset stats for pretrained models * check normalization before applying * add link to quantile augument script * fix(policies): transformers import for ci in PI0 & PI05 (#2039) * fix(policies): transformers import for ci in PI0 * fix(policies): transformers import for ci in PI05 * test(processor): fix expected raise when normalization types are missing (#2040) * switch normalization order pipeline for pi05 * Fix/quantiles script (#2064) * refactor augment stats with quantiles script add parallelization for faster processing shift the quantile normalization between -1 1 * fix replay buffer tests * fix comment * overwrite the pipeline normalization features with the policy features * remove double normalization overwrite * cleanup from pretrained * remove typo * also set norm_map * fix(augment_quantiles) images incorrectly divided by 255 * clamp quantiles * link to lerobot base models * rename tests * encorperate PR feedback * update docstring for RunningQuantileStats * update doc links * Revert "clamp quantiles" This reverts commit 172207471c8f2cb62958e9a9e6a0535ba3ff67d4. * fix self.paligemma * fix tests related to quantiles that were scaled to [0,1], the new range is [-1, 1] * fix libero doc and use different transformer branch * use fix branch instead of feat * update results libero * add new line * fix formatting * precommit * update results libero * update libero doc * update title * final changes * add quantiles to test * run pre commit --------- Signed-off-by: Steven Palma <imstevenpmwork@ieee.org> Co-authored-by: Michel Aractingi <michel.aractingi@huggingface.co> Co-authored-by: Adil Zouitine <adilzouitinegm@gmail.com> Co-authored-by: Steven Palma <imstevenpmwork@ieee.org> Co-authored-by: Steven Palma <steven.palma@huggingface.co>
This commit is contained in:
Pepijn
@@ -166,6 +166,226 @@ def test_min_max_normalization(observation_normalizer):
|
||||
assert torch.allclose(normalized_obs[OBS_STATE], expected_state, atol=1e-6)
|
||||
|
||||
|
||||
def test_quantile_normalization():
|
||||
"""Test QUANTILES mode using 1st-99th percentiles."""
|
||||
features = {
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.STATE: NormalizationMode.QUANTILES,
|
||||
}
|
||||
stats = {
|
||||
"observation.state": {
|
||||
"q01": np.array([0.1, -0.8]), # 1st percentile
|
||||
"q99": np.array([0.9, 0.8]), # 99th percentile
|
||||
},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
observation = {
|
||||
"observation.state": torch.tensor([0.5, 0.0]),
|
||||
}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
normalized_transition = normalizer(transition)
|
||||
normalized_obs = normalized_transition[TransitionKey.OBSERVATION]
|
||||
|
||||
# Check quantile normalization to [-1, 1]
|
||||
# For state[0]: 2 * (0.5 - 0.1) / (0.9 - 0.1) - 1 = 2 * 0.4 / 0.8 - 1 = 0.0
|
||||
# For state[1]: 2 * (0.0 - (-0.8)) / (0.8 - (-0.8)) - 1 = 2 * 0.8 / 1.6 - 1 = 0.0
|
||||
expected_state = torch.tensor([0.0, 0.0])
|
||||
assert torch.allclose(normalized_obs["observation.state"], expected_state, atol=1e-6)
|
||||
|
||||
|
||||
def test_quantile10_normalization():
|
||||
"""Test QUANTILE10 mode using 10th-90th percentiles."""
|
||||
features = {
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.STATE: NormalizationMode.QUANTILE10,
|
||||
}
|
||||
stats = {
|
||||
"observation.state": {
|
||||
"q10": np.array([0.2, -0.6]), # 10th percentile
|
||||
"q90": np.array([0.8, 0.6]), # 90th percentile
|
||||
},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
observation = {
|
||||
"observation.state": torch.tensor([0.5, 0.0]),
|
||||
}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
normalized_transition = normalizer(transition)
|
||||
normalized_obs = normalized_transition[TransitionKey.OBSERVATION]
|
||||
|
||||
# Check quantile normalization to [-1, 1]
|
||||
# For state[0]: 2 * (0.5 - 0.2) / (0.8 - 0.2) - 1 = 2 * 0.3 / 0.6 - 1 = 0.0
|
||||
# For state[1]: 2 * (0.0 - (-0.6)) / (0.6 - (-0.6)) - 1 = 2 * 0.6 / 1.2 - 1 = 0.0
|
||||
expected_state = torch.tensor([0.0, 0.0])
|
||||
assert torch.allclose(normalized_obs["observation.state"], expected_state, atol=1e-6)
|
||||
|
||||
|
||||
def test_quantile_unnormalization():
|
||||
"""Test that quantile normalization can be reversed properly."""
|
||||
features = {
|
||||
"action": PolicyFeature(FeatureType.ACTION, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.ACTION: NormalizationMode.QUANTILES,
|
||||
}
|
||||
stats = {
|
||||
"action": {
|
||||
"q01": np.array([0.1, -0.8]),
|
||||
"q99": np.array([0.9, 0.8]),
|
||||
},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
unnormalizer = UnnormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
# Test round-trip normalization
|
||||
original_action = torch.tensor([0.5, 0.0])
|
||||
transition = create_transition(action=original_action)
|
||||
|
||||
# Normalize then unnormalize
|
||||
normalized = normalizer(transition)
|
||||
unnormalized = unnormalizer(normalized)
|
||||
|
||||
# Should recover original values
|
||||
recovered_action = unnormalized[TransitionKey.ACTION]
|
||||
assert torch.allclose(recovered_action, original_action, atol=1e-6)
|
||||
|
||||
|
||||
def test_quantile_division_by_zero():
|
||||
"""Test quantile normalization handles edge case where q01 == q99."""
|
||||
features = {
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (1,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.STATE: NormalizationMode.QUANTILES,
|
||||
}
|
||||
stats = {
|
||||
"observation.state": {
|
||||
"q01": np.array([0.5]), # Same value
|
||||
"q99": np.array([0.5]), # Same value -> division by zero case
|
||||
},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
observation = {
|
||||
"observation.state": torch.tensor([0.5]),
|
||||
}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
# Should not crash and should handle gracefully
|
||||
normalized_transition = normalizer(transition)
|
||||
normalized_obs = normalized_transition[TransitionKey.OBSERVATION]
|
||||
|
||||
# When quantiles are identical, should normalize to 0 (due to epsilon handling)
|
||||
assert torch.isfinite(normalized_obs["observation.state"]).all()
|
||||
|
||||
|
||||
def test_quantile_partial_stats():
|
||||
"""Test that quantile normalization handles missing quantile stats by raising."""
|
||||
features = {
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.STATE: NormalizationMode.QUANTILES,
|
||||
}
|
||||
|
||||
# Missing q99 - should pass through unchanged
|
||||
stats_partial = {
|
||||
"observation.state": {
|
||||
"q01": np.array([0.1, -0.8]), # Only q01, missing q99
|
||||
},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats_partial)
|
||||
|
||||
observation = {
|
||||
"observation.state": torch.tensor([0.5, 0.0]),
|
||||
}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
with pytest.raises(ValueError, match="QUANTILES normalization mode requires q01 and q99 stats"):
|
||||
_ = normalizer(transition)
|
||||
|
||||
|
||||
def test_quantile_mixed_with_other_modes():
|
||||
"""Test quantile normalization mixed with other normalization modes."""
|
||||
features = {
|
||||
"observation.image": PolicyFeature(FeatureType.VISUAL, (3,)),
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (2,)),
|
||||
"action": PolicyFeature(FeatureType.ACTION, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.VISUAL: NormalizationMode.MEAN_STD, # Standard normalization
|
||||
FeatureType.STATE: NormalizationMode.QUANTILES, # Quantile normalization
|
||||
FeatureType.ACTION: NormalizationMode.QUANTILE10, # Different quantile mode
|
||||
}
|
||||
stats = {
|
||||
"observation.image": {"mean": [0.5, 0.5, 0.5], "std": [0.2, 0.2, 0.2]},
|
||||
"observation.state": {"q01": [0.1, -0.8], "q99": [0.9, 0.8]},
|
||||
"action": {"q10": [0.2, -0.6], "q90": [0.8, 0.6]},
|
||||
}
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
observation = {
|
||||
"observation.image": torch.tensor([0.7, 0.5, 0.3]),
|
||||
"observation.state": torch.tensor([0.5, 0.0]), # Should use QUANTILES
|
||||
}
|
||||
action = torch.tensor([0.5, 0.0]) # Should use QUANTILE10
|
||||
transition = create_transition(observation=observation, action=action)
|
||||
|
||||
normalized_transition = normalizer(transition)
|
||||
normalized_obs = normalized_transition[TransitionKey.OBSERVATION]
|
||||
normalized_action = normalized_transition[TransitionKey.ACTION]
|
||||
|
||||
# Image should be mean/std normalized: (0.7 - 0.5) / 0.2 = 1.0, etc.
|
||||
expected_image = (torch.tensor([0.7, 0.5, 0.3]) - 0.5) / 0.2
|
||||
assert torch.allclose(normalized_obs["observation.image"], expected_image)
|
||||
|
||||
# State should be quantile normalized: 2 * (0.5 - 0.1) / (0.9 - 0.1) - 1 = 0.0, etc.
|
||||
expected_state = torch.tensor([0.0, 0.0])
|
||||
assert torch.allclose(normalized_obs["observation.state"], expected_state, atol=1e-6)
|
||||
|
||||
# Action should be quantile10 normalized: 2 * (0.5 - 0.2) / (0.8 - 0.2) - 1 = 0.0, etc.
|
||||
expected_action = torch.tensor([0.0, 0.0])
|
||||
assert torch.allclose(normalized_action, expected_action, atol=1e-6)
|
||||
|
||||
|
||||
def test_quantile_with_missing_stats():
|
||||
"""Test that quantile normalization handles completely missing stats gracefully."""
|
||||
features = {
|
||||
"observation.state": PolicyFeature(FeatureType.STATE, (2,)),
|
||||
}
|
||||
norm_map = {
|
||||
FeatureType.STATE: NormalizationMode.QUANTILES,
|
||||
}
|
||||
stats = {} # No stats provided
|
||||
|
||||
normalizer = NormalizerProcessorStep(features=features, norm_map=norm_map, stats=stats)
|
||||
|
||||
observation = {
|
||||
"observation.state": torch.tensor([0.5, 0.0]),
|
||||
}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
normalized_transition = normalizer(transition)
|
||||
normalized_obs = normalized_transition[TransitionKey.OBSERVATION]
|
||||
|
||||
# Should pass through unchanged when no stats available
|
||||
assert torch.allclose(normalized_obs["observation.state"], observation["observation.state"])
|
||||
|
||||
|
||||
def test_selective_normalization(observation_stats):
|
||||
features = _create_observation_features()
|
||||
norm_map = _create_observation_norm_map()
|
||||
@@ -547,7 +767,7 @@ def test_empty_stats():
|
||||
|
||||
|
||||
def test_partial_stats():
|
||||
"""If statistics are incomplete, the value should pass through unchanged."""
|
||||
"""If statistics are incomplete, we should raise."""
|
||||
stats = {OBS_IMAGE: {"mean": [0.5]}} # Missing std / (min,max)
|
||||
features = {OBS_IMAGE: PolicyFeature(FeatureType.VISUAL, (3, 96, 96))}
|
||||
norm_map = {FeatureType.VISUAL: NormalizationMode.MEAN_STD}
|
||||
@@ -555,8 +775,8 @@ def test_partial_stats():
|
||||
observation = {OBS_IMAGE: torch.tensor([0.7])}
|
||||
transition = create_transition(observation=observation)
|
||||
|
||||
processed = normalizer(transition)[TransitionKey.OBSERVATION]
|
||||
assert torch.allclose(processed[OBS_IMAGE], observation[OBS_IMAGE])
|
||||
with pytest.raises(ValueError, match="MEAN_STD normalization mode requires mean and std stats"):
|
||||
_ = normalizer(transition)[TransitionKey.OBSERVATION]
|
||||
|
||||
|
||||
def test_missing_action_stats_no_error():
|
||||
|
||||
Reference in New Issue
Block a user