TQDMLoggerBackend

class composer.loggers.tqdm_logger.TQDMLoggerBackend(config: Optional[Dict[str, Any]] = None)

Bases: composer.core.logging.base_backend.RankZeroLoggerBackend

Shows TQDM progress bars.

During training, the progress bar logs the batch and training loss. During validation, the progress bar logs the batch and validation accuracy.

Example output:

Epoch 1: 100%|██████████| 64/64 [00:01<00:00, 53.17it/s, loss/train=2.3023]
Epoch 1 (val): 100%|██████████| 20/20 [00:00<00:00, 100.96it/s, accuracy/val=0.0995]

Note

It is currently not possible to show additional metrics. Custom metrics for the TQDM progress bar will be supported in a future version.

Parameters

config (dict or None, optional) – Trainer configuration. If provided, it is printed to the terminal as YAML.

after_backward(state: State, logger: Logger) → None

Called on the AFTER_BACKWARD event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

epoch_end(state: State, logger: Logger) → None

Called on the EPOCH_END event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

epoch_start(state: State, logger: Logger) → None

Called on the EPOCH_START event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

eval_after_forward(state: State, logger: Logger) → None

Called on the EVAL_AFTER_FORWARD event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

eval_end(state: State, logger: Logger) → None

Called on the EVAL_END event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

eval_start(state: State, logger: Logger) → None

Called on the EVAL_START event.

Parameters

• state (State) – The global state.

• logger (Logger) – The logger.

load_state_dict(state: composer.core.types.StateDict) → None

Restores the state of the object.

Parameters

state (StateDict) – The state of the object, as previously returned by state_dict()

state_dict() → composer.core.types.StateDict

Returns a dictionary representing the internal state.

The returned dictionary must be pickale-able via torch.save().

Returns

StateDict – The state of the object