Return a 7 alpha-numeric character random string. We could use the full uuid() for better uniqueness properties, but it makes the filenames long

For the official ablations on MSRVTT, we provide MoEE with the same hyperparam budget as CE and run a search to find the best hyperparams. For th

Convert json key used for a given datset into the name used in the codebase. Args: json_key: the key to be converted Returns:

Load features from disk

TODO: Samuel - this is redundant due to log_summary() func in log_parser should refactor after deadline.

Parse compact command line strings of the form: --key1 val_a|val_b --key2 val_c|val_d (here a vertical bar represents multiple values)

Recursively convert pathlib objects to strings to enable serialization

Remove checkpoints that are no longer neededself. NOTE: This function assumes that the `best` checkpoint has already been renamed to

Prune audio-based features from the config and dest_path name (necessary for datasets like MSVD which do not possess sound.) If the audio feature

Resets the meter with empty member variables

Catch any nans in the inputs/clusters

Run sanity checks on loaded data

Set seeds for randomisation libraries. Args: seed: the seed value

Setup logging configuration.

Report basic statistics about feature availability and variable lengths across the different subsets of the data.

Convert a string of text into an embedding. Args: text: the content to be embedded Returns: (d x n) array, w

Returns the model's average precision for each class Return: ap (FloatTensor): 1xK tensor, with avg precision for each class k

If visualization is configured to use: return add_data() methods of tensorboard with additional information (step, ta

Initialize the HTML classes Parameters: web_dir (str) -- a directory that stores the webpage. HTML file will be creat

Initialize the Visualizer class Create an HTML object for saveing HTML filters