Update app.py

app.py

@@ -41,20 +41,7 @@ def query(payload):
 #model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B")
 
 
-config = ViTConfig(num_hidden_layers=12, hidden_size=768)
-model = ViTForImageClassification(config)
 
-#print(config)
-
-feature_extractor = ViTFeatureExtractor()
-# or, to load one that corresponds to a checkpoint on the hub:
-#feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
-
-#the following gets called by classify_image() 
-feature_extractor = PerceiverFeatureExtractor.from_pretrained("deepmind/vision-perceiver-conv")
-model = PerceiverForImageClassificationConvProcessing.from_pretrained("deepmind/vision-perceiver-conv")
-#google/vit-base-patch16-224, deepmind/vision-perceiver-conv
-image_pipe = ImageClassificationPipeline(model=model, feature_extractor=feature_extractor)
 
 
 
@@ -168,12 +155,30 @@ def self_caption(image):
   print(pred_dictionary)
   #return(pred_dictionary)
   preds = ' '.join(preds)
-  story = create_story(preds)
-  story = ' '.join(story)
+  #inference(input_sentence, max_length, sample_or_greedy, seed=42)
+  story = inference(preds, 32, "Sample", 42) 
+  #story = create_story(preds)
+  #story = ' '.join(story)
   return story
 
 
 def classify_image(image):
+  config = ViTConfig(num_hidden_layers=12, hidden_size=768)
+  model = ViTForImageClassification(config)
+
+  #print(config)
+
+  feature_extractor = ViTFeatureExtractor()
+  # or, to load one that corresponds to a checkpoint on the hub:
+  #feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
+
+  #the following gets called by classify_image() 
+  feature_extractor = PerceiverFeatureExtractor.from_pretrained("deepmind/vision-perceiver-conv")
+  model = PerceiverForImageClassificationConvProcessing.from_pretrained("deepmind/vision-perceiver-conv")
+  #google/vit-base-patch16-224, deepmind/vision-perceiver-conv
+  image_pipe = ImageClassificationPipeline(model=model, feature_extractor=feature_extractor)
+ 
+  
   results = image_pipe(image)
   
   print("RESULTS")