Plot 3D bar plots for evaluation statistics with (x, y, z) = (n_episodes, denoising_step, metric). Parameters: - eval_statistics_list: L

Plot multiple 3D surface plots in a single figure to compare evaluation statistics across different groups, such as finetuned and pretrained

Plot evaluation statistics for multiple models with filtered denoising steps. Parameters: - eval_statistics_list: List of tuples contain

Plot evaluation statistics for multiple models. Parameters: - eval_statistics_list: List of tuples containing evaluation statistics for

Plot average episode reward vs. denoising steps for multiple models. Parameters: - eval_statistics_list: List of tuples containing evalu

Compute log-density of the prior distribution. Useful for computing the log-likelihood via probability flow ODE. Args: z:

input: torch.Tensor([B, Ta , Da]) log \sigma^2 output: torch.Tensor([B, 1, Ta * Da]), sigma, floating poin

Convert Quaternion to Euler Angles. See rotation.py for notes

Ignore passed-in arguments like seed

Ignore passed-in arguments like seed

Ignore passed-in arguments like seed

Create the reverse-time SDE/ODE. Args: score_fn: A time-dependent score-based model that takes x and t and returns the score.

Collect an expert dataset from an SAC agent in the Humanoid-v3 environment and save it as an .npz file. Parameters: - sac: The train

Sample time steps from a specified distribution in [0, 1). Args: batch_size: Number of time samples to generate. time

saves model to disk; no ema recorded because we are doing RLFT. for evaluation purpose, set ``only_save_policy`` to True. This option

saves model to disk; no ema recorded because we are doing RLFT. for evaluation purpose, set ``only_save_policy`` to True. This option

Not doing validation right now

Show the legends of your curves in desired order. **Use case** ```python handle, = ax.plot(..., label=xxx, ...) handles.app

Takes in a chunk of actions of length n_action_steps and steps the environment n_action_steps times and returns an aggregated

obs_venv: dict containing numpy.ndarray

Do image augmentation on the whole buffer, and compute value, logprobabilities, adv, and return on those augmentated data. obs_venv:

obs_venv: dict containing numpy.ndarray

Do image augmentation on the whole buffer, and compute value, logprobabilities, adv, and return on those augmentated data. obs_venv:

obs_venv: dict containing numpy.ndarray

bug fix

bug fix

bug fix

bug fix

Loads .pkl file generated by training logs and pushes its entries to WandB.

End time of the SDE.

Match input batch size, but do not depend on input

Match input batch size, but do not depend on input

repeat states/images if using history observation at the beginning of the episode