Struct TensorBoardLogger 

pub struct TensorBoardLogger { /* private fields */ }

TensorBoard logger that writes real event files.

Writes TensorBoard event files in the tfevents format, which can be visualized using TensorBoard.

Example

use tensorlogic_train::TensorBoardLogger;
use tensorlogic_train::LoggingBackend;

let mut logger = TensorBoardLogger::new("./logs/run1").unwrap();
logger.log_scalar("loss", 0.5, 1).unwrap();
logger.log_scalar("accuracy", 0.95, 1).unwrap();
logger.flush().unwrap();

Implementations

impl TensorBoardLogger

pub fn new<P: AsRef<Path>>(log_dir: P) -> TrainResult<Self>

Create a new TensorBoard logger.

Arguments

• log_dir - Directory for TensorBoard logs

pub fn log_dir(&self) -> &Path

Get the log directory.

pub fn file_path(&self) -> &Path

Get the event file path.

pub fn log_histogram( &mut self, tag: &str, values: &[f64], step: usize, ) -> TrainResult<()>

Log a histogram (weight distributions).

Trait Implementations

impl Debug for TensorBoardLogger

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl LoggingBackend for TensorBoardLogger

fn log_scalar(&mut self, name: &str, value: f64, step: usize) -> TrainResult<()>

Log a scalar metric.

fn log_text(&mut self, message: &str) -> TrainResult<()>

Log a text message.

fn flush(&mut self) -> TrainResult<()>

Flush any buffered logs.