using /tmp, should be writable

vit_captioning/generate.py

@@ -24,23 +24,23 @@ class CaptionGenerator:
             print("No GPU found, falling back to CPU.")
 
         # Load tokenizer
-        self.tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')
+        #self.tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')
         #HF needs all model downloads to a special read-write cache dir
-        #self.tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased', cache_dir="/data")
+        self.tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased', cache_dir="/tmp")
 
         # Select encoder, processor, output dim
         if model_type == "ViTEncoder":
             self.encoder = ViTEncoder().to(self.device)
             self.encoder_dim = 768
-            self.processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k")
+            #self.processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k")
             #HF needs all model downloads to a special read-write cache dir
-            #self.processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k", cache_dir="/data")
+            self.processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k", cache_dir="/tmp")
         elif model_type == "CLIPEncoder":
             self.encoder = CLIPEncoder().to(self.device)
             self.encoder_dim = 512
-            self.processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
+            #self.processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
             #HF needs all model downloads to a special read-write cache dir
-            #self.processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32", cache_dir="/data")
+            self.processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32", cache_dir="/tmp")
         else:
             raise ValueError("Unknown model type")