Launch an ML training experiment on tracebloc in under 10 minutes. Connect your account, upload a model, link it to a dataset, configure training parameters, and start benchmarking — all from a single notebook.
No local setup. Click the badge above or:
Copy the notebook to your Drive and start running cells.
git clone https://github.com/tracebloc/quickstart.git
cd quickstart
# Pick the extra that matches your ML framework:
pip install "tracebloc[pytorch]>=1.2.42" # most common
# pip install "tracebloc[sklearn]>=1.2.42" # scikit-learn / boosting
# pip install "tracebloc[all]>=1.2.42" # everything
jupyter notebook notebooks/traceblocTrainingGuide.ipynbWhich Pythons work: whatever the SDK's own package metadata declares — see
tracebloc on PyPI. This README
deliberately does not repeat the range; a copy here would go stale against the
package, which is exactly the failure this notebook was fixed for.
If pip answers No matching distribution found for tracebloc, that most often
means your interpreter is outside that range rather than the package being
missing — but it can also mean an unreachable index or a custom --index-url.
The install cell prints pip's own answer either way: on failure it shows the
range pip actually read alongside the Python you are on. On macOS the default
python3 is frequently ahead of the supported range, so check
python3 --version first when running locally.
TensorFlow uploads were removed in SDK 1.0.0, so there is no [tensorflow]
extra — the extras are [pytorch], [sklearn], [catboost], [lightgbm],
[xgboost], [lifelines], [scikit-survival] and [all].
| Step | What you do |
|---|---|
| 1 | Connect to tracebloc with your email + password |
| 2 | Upload a model from the model zoo or your own |
| 3 | Link it to a dataset from your use case |
| 4 | Configure training — epochs, batch size, learning rate, augmentation |
| 5 | Start training — model runs inside your secure Kubernetes environment |
Results appear on the use case leaderboard in the tracebloc web app.
- A tracebloc account — sign up free
- An active use case with a dataset — how to join one
- A model file — grab one from the model zoo or build your own
Platform · Docs · Model zoo · PyPI package · Discord
Apache 2.0 — see LICENSE.
Need help? support@tracebloc.io or open an issue.