We Raised $8M Series A to Continue Building Experiment Tracking and Model Registry That “Just Works”
Neptune vs TensorBoard
vs
Commercial Requirements
Standalone component or a part of a broader ML platform?
Standalone component. ML metadata store that focuses on experiment tracking and model registry
Open source tool which is a part of the TensorFlow ecosystem
Is the product available on-premises and / or in your private/public cloud?
Is the product delivered as commercial software, open-source software, or a managed cloud service?
Managed cloud service
TensorBoard is open-source, while TensorBoard.dev is available as a free managed cloud service
SLOs / SLAs: Does the vendor provide guarantees around service levels?
Email and chat support for individuals and teams. Well defined support SLAs for enterprise customers
Community support only
General Capabilities
What are the infrastructure requirements?
No special requirements other than having the neptune-client installed and access to the internet if using managed hosting. Check here for infrastructure requirements for on-prem deployment
Basic logging can be done by having just TensorBoard installed. However, most advanced logging also requires TensorFlow to be installed
How much do you have to change in your training process?
Minimal. Just a few lines of code needed for traking. Read more
Minimal if already using the TensorFlow framework, else significant
Does it integrate with the training process via CLI/YAML/Client library?
Yes, through the neptune-client library
TensorBoard is available both as a client library and CLI. TensorBoard .dev is available only as a CLI
Serverless UI
No
Yes
How can you access model metadata?
– gRPC API
No
No
– REST API
No
No
– Java SDK
No
No
– Julia SDK
No
No
Supported operations
Logging modes
Mobile support
No
No
Webhooks and notifications
No
No
Experiment Tracking
Dataset
– Dataset slicing support
No
No
Code versions
Metrics and losses
Rich format
– Prediction visualization (tabular)
No
No
– Prediction visualization (image)
No
No
Hardware consumption
– TPU
No
No
System information
– Execution command
No
No
Environment config
Files
– External file reference (s3 buckets)
No
No
Parameters and metrics
– Parameter Importance plot
No
No
– Slice plot
No
No
– EDF plot
No
No
Rich format (side by side)
– Video
No
No
– Audio
No
No
– Plots
No
No
– Interactive visualization (HTML)
No
No
– Prediction visualization (image, video, audio)
No
No
Code
– Git
No
No
– Source files
No
No
– Notebooks
Yes
No
Environment
– pip requirements.txt
No
No
– conda env.yml
No
No
– Docker Dockerfile
No
No
Hardware
System information
– Execution command
No
No
Data versions
– Dataset diff
No
No
– External reference version diff (s3)
No
No
Files
– Models
No
No
– CSV
No
No
Custom compare dashboards
Experiment table customization
– Renaming columns in the UI
Yes
No
– Adding colors to columns
Yes
No
Experiment filtering and searching
– Query language vs fixed selectors
Query language
Regex with limited query language on the TensorBoard.dev experiments homepage
Custom dashboards for a single experiment
– Custom widgets and plugins
Nos
No
One-command experiment re-run
No
No
Experiment lineage
– List of datasets used downstream
No
No
– List of other artifacts (models) used downstream
No
No
– Downstream artifact dependency graph
No
No
Reproducibility protocol
Limited
No
Saving/fetching/caching datasets for experiments
No
No
Interactive project-level reports
No
No
Model Registry
Environment versions
No
No
Model files (packaged models, model weights, pointers to artifact storage)
No
No
Models/experiments created downstream
No
No
History of evaluation/testing runs
No
No
Support for continuous testing
No
No
Users who created a model or downstream experiments
No
No
Main stage transition tags (develop, stage, production)
No
No
Custom stage tags
No
No
Locking model version and downstream runs, experiments, and artifacts
No
No
Adding annotations/comments and approvals from the UI
No
No
Model compare (current vs challenger etc)
No
No
Compatibility audit (input/output schema)
No
No
Compliance audit (datasets used, creation process approvals, results/explanations approvals)
No
No
Webhooks
No
No
Model accessibility
No
No
Support for continuous testing
No
No
Imntegrations with CI/CD tools
No
No
Registered models
No
No
Active models
No
No
By metadata/artifacts used to create it
No
No
By date
No
No
By user/owner
No
No
By production stage
No
No
Search query language
No
No
Native packaging system
No
No
Compatibility with packaging protocols (ONNX, etc)
No
No
One model one file or flexible structure
No
No
Integrations and Support
FBProphet
No
No
Paddle
No
No
Spacy
No
No
Spark MLlib
No
No
Statsmodel
No
No
Hyperopt
No
No
This page was updated on 8 August 2021. Some information may be outdated.
Report outdated information here.
What are the key advantages of Neptune, then?
- Easy, one-time setup to have all metadata in one place
- Intuitive, out-of-the box runs comparison
- Hosted and sharable environment
- User management and team collaboration features
- Scalability with thousands of runs
See these features in action
1. Create a free account
Sign up2. Install Neptune client library
pip install neptune-client
3. Add logging to your script
import neptune.new as neptune
run = neptune.init('Me/MyProject')
run['params'] = {'lr':0.1, 'dropout':0.4}
run['test_accuracy'] = 0.84
4. Or see how it works in a notebook (no registration)
Try live notebookAlready using TensorBoard?
Convert your TensorBoard logs directory with Neptune:
neptune tensorboard --project USER_NAME/PROJECT_NAME
Or connect a new project in 5 minutes.
You can use a cloud-hosted service or have Neptune on your own servers.
Want to know why others switched to Neptune?
InstaDeep, an EMEA leader in delivering decision-making AI products, switched from TensorBoard to Neptune.
Read about the biggest challenges they faced with TensorBoard, and why Neptune helped them manage the experimentation process better.
Thousands of ML people already chose their tool
“Previously used tensorboard and azureml but Neptune is hugely better. In particular, getting started is really easy; documentation is excellent, and the layout of charts and parameters is much clearer.”
AI Engineer and Data Scientist
“I’d say the advantage (of TensorBoard) is that it’s free and it works pretty well but anytime an engineer wanted to show the team some training curve, they’d need to start the VM (Virtual Machine) containing the logs, or make their localhost port available, expose it to the internet, it was not very secure… When you end up having to start a VM just to visualize some logs, you realize there should be a better tool.”
DeepChain and BioAI Lead at InstaDeep
“This thing is so much better than Tensorboard, love you guys for creating it!”
Junior Researcher at Rényi Alfréd Institute of Mathematics in Budapest, Hungary
“While logging experiments is great, what sets Neptune apart for us at the lab is the ease of sharing those logs. The ability to just send a Neptune link in slack and letting my coworkers see the results for themselves is awesome. Previously, we used Tensorboard + locally saved CSVs and would have to send screenshots and CSV files back and forth which would easily get lost. So I’d say Neptune’s ability to facilitate collaboration is the biggest plus.”
Computer Science Undergraduate at Saint Louis University
“(…) thanks for the great tool, has been really useful for keeping track of the experiments for my Master’s thesis. Way better than the other tools I’ve tried (comet / wandb).
I guess the main reason I prefer neptune is the interface, it is the cleanest and most intuitive in my opinion, the table in the center view just makes a great deal of sense. I like that it’s possible to set up and save the different view configurations as well. Also, the comparison is not as clunky as for instance with wandb. Another plus is the integration with ignite, as that’s what I’m using as the high-level framework for model training.”
Data Science and AI student, Kranenburg, Germany
“For me the most important thing about Neptune is its flexibility. Even if I’m training with Keras or Tensorflow on my local laptop, and my colleagues are using fast.ai on a virtual machine, we can share our results in a common environment.”
Senior NLP/ML Engineer
