Visualization with TensorBoard

on YouTube ↗05m

" summary

PyTorch exposes native TensorBoard support since PyTorch 1.2 and uses a SummaryWriter that logs to local or remote directories (S3) so training curves, QPS and CPU throughput appear via add_scalar/add_image/add_histogram APIs. Model graph, embedding projector and parallel-coordinate hyperparameter views run inside TensorBoard; the What‑If tool integrates with a custom predict function and TorchServe for per‑segment fairness checks. Upcoming profiler shows per‑operator and GPU‑kernel timings; plotly and alert plugins add interactive plots and thresholded alerts.

AI-written summary · about

" chapters (3)

1. 0:00Introduction
2. 2:47Model Visualization
3. 4:13Features

" related talks

• 12mPyTorch Partner CollaborationsGeeta Chauhan · PyTorch Ecosystem Day 2021
• 42mHigh Performance Distributed TensorflowChris Fregly · PyData Seattle 2017
• 14mSkyline: Interactive In-Editor Computational Performance Profiling for Deep Neural Network TrainingGennady Pekhimenko, Geoffrey X. Yu, Tovi Grossman · UIST 2020
• 43mExploring Deep Learning Framework PyTorchStephanie Kim · PyCon 2018
• 36mHigh-Performance Distributed TensorflowChris Fregly · PyData London 2017
• 05mInterviewChris Fregly · AI By the Bay 2017
• 26mThunder: Distribute and Optimize Your PyTorch Models With Zero...Luca Antiga, Thomas Viehmann · PyTorch Conference 2025
• 38mThe Hitchhiker's Guide to Tensorflow: from Tensors to running models in productionSergii Khomenko · GDG DevFest Ukraine 2017