Instructions to use Namdn/tf-savedmodel-native-ops-poc with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- TF-Keras
How to use Namdn/tf-savedmodel-native-ops-poc with TF-Keras:
# Note: 'keras<3.x' or 'tf_keras' must be installed (legacy) # See https://github.com/keras-team/tf-keras for more details. from huggingface_hub import from_pretrained_keras model = from_pretrained_keras("Namdn/tf-savedmodel-native-ops-poc") - Notebooks
- Google Colab
- Kaggle
| license: mit | |
| tags: | |
| - tensorflow | |
| - savedmodel | |
| - security-research | |
| # TF SavedModel Native Ops PoC | |
| Security research — demonstrates arbitrary file write via native TensorFlow ops (`tf.io.write_file`) embedded in SavedModel graph. | |
| **DO NOT LOAD THIS MODEL IN PRODUCTION.** | |
| ## Vector | |
| - `tf.io.write_file` is a native C++ kernel op | |
| - Embedded directly in the computation graph | |
| - Invisible to Python-level security scanners | |
| - Executes when model inference is called | |
| ## Usage (for testing only) | |
| ```python | |
| import tensorflow as tf | |
| model = tf.saved_model.load("./") | |
| # Calling model triggers file write | |
| result = model(tf.random.normal([1, 10])) | |
| ``` | |