generated: '2026-06-20'
method: searched
source: https://www.tensorflow.org/guide/saved_model, https://www.tensorflow.org/tensorboard, https://www.tensorflow.org/js/guide/conversion, https://www.tensorflow.org/lite/models/convert
note: >-
  TensorFlow ships several first-party command-line tools rather than one unified CLI. Each is
  installed with its corresponding package (see packages/tensorflow-packages.yml).
tools:
  - name: saved_model_cli
    package: tensorflow (PyPI)
    description: Inspect and run SavedModels — list signatures, tensor shapes/dtypes, and execute a signature.
    key_commands:
      - "saved_model_cli show --dir <path> --all"
      - "saved_model_cli run --dir <path> --tag_set serve --signature_def serving_default --input_exprs '...'"
  - name: tensorboard
    package: tensorboard (PyPI)
    description: Launch the TensorBoard visualization web server against a logdir.
    key_commands:
      - "tensorboard --logdir <dir>"
      - "tensorboard --logdir <dir> --port 6006"
  - name: tensorflowjs_converter
    package: tensorflowjs (PyPI)
    description: Convert TensorFlow SavedModels/Keras models to the TensorFlow.js web format.
    key_commands:
      - "tensorflowjs_converter --input_format=tf_saved_model <src> <dest>"
  - name: tflite_convert
    package: tensorflow (PyPI)
    description: Convert TensorFlow models to the TensorFlow Lite (.tflite) format for on-device inference.
    key_commands:
      - "tflite_convert --saved_model_dir=<dir> --output_file=<model.tflite>"