Say hello to more sophisticated experiment tracking
You have great visualization options for your research projects.
But Tensorboard’s other features are limited.
Add Neptune and gain the ability to:
- Collaborate smoothly (no screenshots required)
- Track experiments at scale (without the screen lags)
- Compare results with ease (bye-bye cluttered UI)
Trusted by researchers at enterprises & universities training models at scale
Advanced experiment tracking features
More metadata. More precision. More possibilities.
You love Tensorboard’s charts and graphs. You’d also love more features to accurately track and compare your model training.
Get the best of both worlds by using Neptune as well.
Including:
Comparison charts
See the metrics of 100s of experiments in a single graph.
Forking of runs
Visualize the whole tree of restared or branched runs out of the box.
Support for more metadata types
Even the ones Tensorboard doesn’t have, like files, videos or system info.
And much more…
Painless collaboration
Stop sharing work with so many screenshots Start creating custom reports in a click
“What sets Neptune apart for us is the ease of sharing logs. The ability to send a Neptune link in Slack and let my coworkers see the results for themselves is awesome.”
– Greg Rolwes, CS Undergrad at Saint Louis University
- Create custom reports in seconds
- Share results in real time
Fast UI at scale
Will scale. Won’t fail.
“We originally used Tensorboard and had scalability issues with many runs. With Neptune, we can go through 1,000 model designs and everything seems to work fine, work well, right out of the box.”
– Brian Geier, Data Scientist at TENET 3
- Instant data visualization
- Easily search & filter data
Take a deep dive into
what makes Neptune different
Neptune
TensorBoard
Commercial offering
Managed cloud service
Open-source
User based and usage based (ingestion data points)
Free
General information
Yes
No
Yes
Yes
Minimal setup—install the Python client and ensure internet access (for managed hosting). Self-hosting requires additional infrastructure; see requirements here.
Basic logging can be done by having just TensorBoard installed. However, most advanced logging also requires TensorFlow to be installed.
A few lines of code via the Neptune Python library.
A few lines of code via client library and CLI.
CLI/custom API and Python SDK
Python SDK, R SDK (limited), Julia SDK (limited)
Search, Update, Delete, Download
Search, Delete, Download
Offline, Disabled/Off, Asynchronous
Offline, Asynchronous
Yes
No
Yes
Yes
No
No
Capabilities
Yes
No
Images, Plots, Video, Audio
Plots, Audio
No
Limited
Console logs, Execution command
No
No
No
No
No
Regex with limited query language
No
No
No
No
No
No
Report outdated information here.