Adapt CLIP to your images and captions.
What it does
Fine-tunes and evaluates CLIP models and produces image-text embeddings, ONNX exports, and TensorRT engines.
When to use it
Use for individual CLIP training, classification, embedding, or deployment actions, rather than an iterative retraining-and-evaluation loop.
How to use it
Give it a CLIP job request and the applicable datasets or model artifacts; it runs the requested action and saves its results.
What you provide
- Your files
Uses
Latest review
What you need · 4
Requires Docker for container-based CLIP actions.
Requires NVIDIA Container Toolkit alongside Docker.
Training uses GPUs; the guidance calls for 16GB+ VRAM per GPU for small, fixed-resolution runs without preservation regularization.
Access · 1
This skill
wandb
Write
Logs runs to wandb