tensorboard#

Source code: tianshou/utils/logger/tensorboard.py

class TensorboardLogger(writer: SummaryWriter, training_interval: int = 1000, test_interval: int = 1, update_interval: int = 1000, info_interval: int = 1, save_interval: int | None = None, write_flush: bool = True)[source]#

Bases: BaseLogger

A logger that relies on tensorboard SummaryWriter by default to visualize and log statistics.

Parameters:

  • writer (SummaryWriter) – the writer to log data.

  • training_interval – the interval size (in env steps) after which log_training_data() will be called.

  • test_interval – the interval size (in env steps) after which log_test_data() will be called.

  • update_interval – the interval size (in env steps) after which log_update_data() will be called.

  • info_interval – the interval size (in env steps) after which the method log_info() will be called.

  • save_interval – the interval size (in env steps) after which the checkpoint and end of epoch related logs will be saved.

  • write_flush – whether to flush tensorboard result after each add_scalar operation. Default to True.

prepare_dict_for_logging(input_dict: dict[str, Any], parent_key: str = '', delimiter: str = '/', exclude_arrays: bool = True) → dict[str, int | Number | number | ndarray | float][source]#

Flattens and filters a nested dictionary by recursively traversing all levels and compressing the keys.

Filtering is performed with respect to valid logging data types.

Parameters:

  • input_dict – The nested dictionary to be flattened and filtered.

  • parent_key – The parent key used as a prefix before the input_dict keys.

  • delimiter – The delimiter used to separate the keys.

  • exclude_arrays – Whether to exclude numpy arrays from the output.

Returns:

A flattened dictionary where the keys are compressed and values are filtered.

write(step_type: str, step: int, data: dict[str, Any]) → None[source]#

Specify how the writer is used to log data.

Parameters:

  • step_type (str) – namespace which the data dict belongs to.

  • step – stands for the ordinate of the data dict.

  • data – the data to write with format {key: value}.

finalize() → None[source]#

Finalize the logger, e.g., close writers and connections.

save_data(epoch: int, env_step: int, update_step: int, save_checkpoint_fn: Callable[[int, int, int], str] | None = None) → None[source]#

Use writer to log metadata when calling save_checkpoint_fn in trainer.

Parameters:

  • epoch – the epoch in trainer.

  • env_step – the env_step in trainer.

  • update_step – the update step count in the trainer.

  • save_checkpoint_fn (function) – a hook defined by user, see trainer documentation for detail.

restore_data() → tuple[int, int, int][source]#

Restore internal data if present and return the metadata from existing log for continuation of training.

If it finds nothing or an error occurs during the recover process, it will return the default parameters.

Returns:

epoch, env_step, update_step.

static restore_logged_data(log_path: str) → dict[str, ndarray | dict[str, dict[str, ndarray | dict[str, TRestoredData]]]][source]#

Restores the logged data from the tensorboard log directory.

The result is a nested dictionary where the keys are the tensorboard keys and the values are the corresponding numpy arrays. The keys in each level form a nested structure, where the hierarchy is represented by the slashes in the tensorboard key-strings.