serialization

This is summary documentation for hippynn.experiment.serialization module. Click here for full documentation.

Links to Full Documentation:

Checkpoint and state generation.

As a user, in most cases you will only need the load functions here.

Functions

check_mapping_devices(map_location, model_device)	Check options for restarting across devices.
create_state(model, controller, metric_tracker)	Create an experiment state dictionary.
create_structure_file(training_modules, ...)	Save an experiment structure.
load_checkpoint(structure_fname, state_fname)	Load checkpoint file from given filename.
load_checkpoint_from_cwd([map_location, ...])	Same as load_checkpoint, but using default filenames.
load_model_from_cwd([map_location, ...])	Only load model from current working directory.
load_saved_tensors(structure_fname, state_fname)	Load torch tensors from file.
restore_checkpoint(structure, state[, ...])	This function loads the parameters from the state dictionary into the modules, optionally tries to restart the database, and sets the RNG state.