AI Blocks › Track your numbers
Training curve viewer
Give it your training code. Get live charts of loss and accuracy, plus a way to compare runs.
Training curve viewer is a free AI skill for the job "Track your numbers". Give the instructions to an AI assistant that supports the required files and tools, then review its output.
Free · No account · MIT licence · Made by Orchestra Research · Thinking: low
Made by Orchestra Research. See the original on GitHub; we copied it as it is, with its licence.
What it does
Live charts of loss and accuracy while your model trains, with runs lined up so you can see which settings worked. You stop guessing from printed numbers and watch the curves. The skill adds TensorBoard logging to your PyTorch or TensorFlow code. It writes scalars for loss and accuracy, adds histograms to spot dead or exploding weights, and logs images, model graphs and embeddings when you want them. Each run goes in its own folder so you can compare them on one page. It also covers hyperparameter tracking and the profiler for slow steps. You start the viewer with one command and open it in your browser. It needs Python and TensorBoard installed.
You give it
- Your training script
- The framework you use, PyTorch or TensorFlow
- What you want to watch, such as loss or images (optional)
- Names for the runs you want to compare (optional)
You get
- Logging code added to your script
- Live charts of training metrics
- A run comparison setup
- The command to launch the viewer
Use it when
- You train a model and want to see loss curves instead of console text.
- You are comparing runs with different learning rates or layer sizes.
- Your model stops improving and you want to inspect weights and gradients.
- Training is slow and you need to find the bottleneck.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.
- It works best when your AI can run Python code and open a browser.