| from typing import Dict |
| import torch |
| import numpy as np |
| import copy |
| import pathlib |
| from diffusion_policy.common.pytorch_util import dict_apply |
| from diffusion_policy.common.replay_buffer import ReplayBuffer |
| from diffusion_policy.common.sampler import SequenceSampler, get_val_mask |
| from diffusion_policy.model.common.normalizer import LinearNormalizer, SingleFieldLinearNormalizer |
| from diffusion_policy.dataset.base_dataset import BaseLowdimDataset |
|
|
| class KitchenLowdimDataset(BaseLowdimDataset): |
| def __init__(self, |
| dataset_dir, |
| horizon=1, |
| pad_before=0, |
| pad_after=0, |
| seed=42, |
| val_ratio=0.0 |
| ): |
| super().__init__() |
|
|
| data_directory = pathlib.Path(dataset_dir) |
| observations = np.load(data_directory / "observations_seq.npy") |
| actions = np.load(data_directory / "actions_seq.npy") |
| masks = np.load(data_directory / "existence_mask.npy") |
|
|
| self.replay_buffer = ReplayBuffer.create_empty_numpy() |
| for i in range(len(masks)): |
| eps_len = int(masks[i].sum()) |
| obs = observations[i,:eps_len].astype(np.float32) |
| action = actions[i,:eps_len].astype(np.float32) |
| data = { |
| 'obs': obs, |
| 'action': action |
| } |
| self.replay_buffer.add_episode(data) |
| |
| val_mask = get_val_mask( |
| n_episodes=self.replay_buffer.n_episodes, |
| val_ratio=val_ratio, |
| seed=seed) |
| train_mask = ~val_mask |
| self.sampler = SequenceSampler( |
| replay_buffer=self.replay_buffer, |
| sequence_length=horizon, |
| pad_before=pad_before, |
| pad_after=pad_after, |
| episode_mask=train_mask) |
|
|
| self.train_mask = train_mask |
| self.horizon = horizon |
| self.pad_before = pad_before |
| self.pad_after = pad_after |
|
|
| def get_validation_dataset(self): |
| val_set = copy.copy(self) |
| val_set.sampler = SequenceSampler( |
| replay_buffer=self.replay_buffer, |
| sequence_length=self.horizon, |
| pad_before=self.pad_before, |
| pad_after=self.pad_after, |
| episode_mask=~self.train_mask |
| ) |
| val_set.train_mask = ~self.train_mask |
| return val_set |
|
|
| def get_normalizer(self, mode='limits', **kwargs): |
| data = { |
| 'obs': self.replay_buffer['obs'], |
| 'action': self.replay_buffer['action'] |
| } |
| if 'range_eps' not in kwargs: |
| |
| kwargs['range_eps'] = 5e-2 |
| normalizer = LinearNormalizer() |
| normalizer.fit(data=data, last_n_dims=1, mode=mode, **kwargs) |
| return normalizer |
|
|
| def get_all_actions(self) -> torch.Tensor: |
| return torch.from_numpy(self.replay_buffer['action']) |
|
|
| def __len__(self) -> int: |
| return len(self.sampler) |
|
|
| def __getitem__(self, idx: int) -> Dict[str, torch.Tensor]: |
| sample = self.sampler.sample_sequence(idx) |
| data = sample |
|
|
| torch_data = dict_apply(data, torch.from_numpy) |
| return torch_data |
|
|
