Add CLIP to the model zoo

docs/scripts/make_model_zoo_docs.py

@@ -111,6 +111,19 @@
     dataset.apply_model(model, label_field="predictions")
 
     session = fo.launch_app(dataset)
+
+{% if 'zero-shot' in tags %}
+    # Make zero-shot label predictions with custom class labels
+
+    custom_labels = ["person", "dog", "cat", "bird", "car", "tree", "chair"]  # can be any list of strings 
+    text_prompt = "A photo of a"  # will be prepended to each class label
+
+    model = foz.load_zoo_model("{{ name }}", class_labels=custom_labels, text_prompt=text_prompt)
+
+    dataset.apply_model(model, label_field="predictions")
+
+    session = fo.launch_app(dataset)
+{% endif %}
 """
 
 

fiftyone/utils/clip/__init__.py

@@ -0,0 +1,6 @@
+"""CLIP model utils.
+"""
+
+from .model import build_model
+from .tokenizer import SimpleTokenizer
+from .zoo import TorchCLIPModelConfig, TorchCLIPModel