12. Troubleshooting

Solutions to common problems when training, compiling and deploying with Tiny ML ModelZoo 1.5.0, and answers to frequently asked questions.

Contents

• 12.1. Common Errors
  • 12.1.1. Setup and configuration
  • 12.1.2. Training
  • 12.1.3. Compilation and model design
  • 12.1.4. Platform
  • 12.1.5. Other tips
• 12.2. FAQ
  • 12.2.1. General
  • 12.2.2. Installation
  • 12.2.3. Models and training
  • 12.2.4. Deployment

12.3. Quick checks

Most problems fall into one of these groups. Start with the matching entry in Common Errors.

Is your Python version 3.10?

Training works on Python 3.10 to 3.14, but compilation needs the TI neural network compiler (ti_mcu_nnc), which is only installed on Python 3.10. See Python 3.11 or newer cannot compile.

python --version

Is the model name valid for this release?

training.model_name must be one of the models listed in Model Zoo Reference. See Unknown model name.

Compilation fails but training works

Check that the TI compiler for your device is installed under TOOLS_PATH (default ~/ti) and that the model follows the NPU layer rules. See Compiler tools not found (TOOLS_PATH) and NPU layer constraint violations.

Outputs do not match on the device

See Golden vector mismatch on the device and Deploying a ModelZoo Run to an SDK Example (Golden-Vector Check).

12.4. Getting help

• Read the log files in the run folder, ./data/projects/<project_name>/run/<date-time>/<model_name>/. See Understanding the Run Output.

• Check the FAQ.

• Report issues at https://github.com/TexasInstruments/tinyml-modelzoo/issues, with the YAML you used and the log excerpt.