Update app.py

app.py

@@ -2,23 +2,45 @@ import torch
 import re 
 import gradio as gr
 from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel
-
+'''
 device='cpu'
 encoder_checkpoint = "Thibalte/captionning_project"
 decoder_checkpoint = "Thibalte/captionning_project"
 model_checkpoint = "Thibalte/captionning_project"
+feature_extractor= ViTImageProcessor.from_pretrained(model_path)
 feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
 tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)
 model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)
+'''
+# Load the trained model
+model_path = "./image-captioning-output"
+
+# Load tokenizer
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+
+#Load ImageProcessor
+feature_extractor= ViTImageProcessor.from_pretrained(model_path)
+
+# Load model
+model = VisionEncoderDecoderModel.from_pretrained(model_path)
+pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
+# use GPT2's eos_token as the pad as well as eos token
+
+
+# generation
+
+
+
+
+print(captions)
 
 
 def predict(image,max_length=24, num_beams=4):
-  image = image.convert('RGB')
-  image = feature_extractor(image, return_tensors="pt").pixel_values.to(device)
-  clean_text = lambda x: x.replace('<|endoftext|>','').split('\n')[0]
-  caption_ids = model.generate(image, max_length = max_length)[0]
-  caption_text = clean_text(tokenizer.decode(caption_ids))
-  return caption_text 
+    image = image.convert('RGB')
+    sequences = model.generate(pixel_values, num_beams=4, max_length=25)
+    sequences = model.generate(pixel_values, num_beams=4, max_length=25)
+    captions = tokenizer.batch_decode(sequences, skip_special_tokens=True)
+  return caption
 
 
 # Gradio Interface