Create generate_embeddings.py

generate_embeddings.py

@@ -0,0 +1,18 @@
+try:
+    embeddings_df = pd.read_pickle('image_embeddings.pickle')
+    index = embeddings_df.shape[0]
+except:
+    index=0
+    embeddings_df = pd.DataFrame(columns=['image_embedding'])
+formats = []
+while  index<tasks_df.shape[0]:
+    image = load_image(tasks_df['image_path'][index])
+    inputs = processor(images=[image], return_tensors="pt").to(model.device)
+    
+    with torch.no_grad():
+        image_embeddings = model.get_image_features(**inputs)    
+    new_row = {'image_embedding': image_embeddings}
+    embeddings_df = pd.concat([embeddings_df, pd.DataFrame([new_row])], ignore_index=True)
+    if index % 100==0:
+         embeddings_df.to_pickle('image_embeddings.pickle')
+    index+=1