Instructions to use FrankLin00/mmrh-ace-03-keras-import-module with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Keras
How to use FrankLin00/mmrh-ace-03-keras-import-module with Keras:
# !pip install -U keras tensorflow huggingface_hub # Keras needs TensorFlow installed to read "hf://" paths, so the tensorflow backend is selected here; # "jax" and "torch" also work for computation once TensorFlow is installed. import os os.environ["KERAS_BACKEND"] = "tensorflow" import keras model = keras.saving.load_model("hf://FrankLin00/mmrh-ace-03-keras-import-module") - Notebooks
- Google Colab
- Kaggle
keras/02-keras-import-module-ace
Inert scanner specimen for academic security research. It preserves the finding's model-metadata carrier: .keras config.json module (ACE). It must not be used in production.
Any active probe directly executes only whoami (argv form, never shell=True) and writes
its output to HFSCAN_KERAS__02_KERAS_IMPORT_MODULE_ACE under /tmp or $TMPDIR; it performs no credential access or
network operation. See the collection-level
MANIFEST.json for hashes and the source finding path.
AFW specimens use the literal inert payload PWNED-INERT. When the vulnerable primitive
forces a container or suffix (for example NumPy .npy, a checkpoint, or .jinja), the
manifest records that format constraint instead of claiming raw-byte control.
- Downloads last month
- 31