You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Fine-tuned OpenCLIP ViT-B/16 for CIFAR-10

This private checkpoint is the frozen classifier used to construct the paired PGD dataset and generate explanation maps in the XAI Attack Detection project.

Training

Base model: OpenCLIP ViT-B-16
Pretrained weights: laion2b_s34b_b88k
Data: all 50,000 CIFAR-10 training images
Objective: ten-way class-prompt similarity cross entropy
Trainable parameters: final two vision transformer blocks and logit scale
Epochs: 3
Learning rate: 1e-5

The uploaded artifact includes Safetensors weights, class prompts/text embeddings, resolved configuration, dependency versions, evaluation metrics, and checksums.

Intended use and limitations

Use this exact revision to reproduce the associated adversarial dataset and explanation maps. It is an experimental research checkpoint, not a production classifier.

License and access

This artifact is private and currently has no reuse license. Base-model and CIFAR-10 terms also apply and must be reviewed before public release.

Downloads last month: -

Safetensors

Model size

0.1B params

Tensor type

F32