utils.writer.tensorboard

Tensorboard writer module.

Tensorboard Objects#

class Tensorboard(Writer)

This class implements a wrapper on Tensorboard's SummaryWriter for logging stats to Tensorboard. Please look here for detailed information on each function: https://pytorch.org/docs/stable/tensorboard.html# Visit this blog for more examples on logging stats to Tensorboard: https://krishansubudhi.github.io/deeplearning/2020/03/24/tensorboard-pytorch.html

def log_scalar(k, v, step)

Log metric to Tensorboard graph.

def log_multi(k, v, step)

Log multiple metrics in the same Tensorboard graph.

def log_model(flat_weights, flat_grads, step)

Log model weights and gradients to Tensorboard.

def log_embedding(tag, mat, labels, step)

Log model embeddings to Tensorboard.

def log_graph(model, device)

Logs model graphs to Tensorboard.

Arguments:

model object - unwrapped model with a function get_sample_input() implemented.
device str - device type.

def log_image(k, v, step, dataformats='HW')

Log image in Tensorboard.

def log_pr_curve(k, preds, labels, step)

Log Precision Recall curve in Tensorboard.

def log_args(args)

Log all the Arguments used in the experiment to Tensorboard.

def log_histogram(param_name, vals, step)

Log histograms in Tensorboard. Avoid using small step size since it impact training time.

def flush()

Flush the SummaryWriter to write out Summary to Tensorboard.

def finish()

Flush the SummaryWriter to write out Summary to Tensorboard and close SummaryWriter.