Creating captioning pipeline with nucleus sampling

pipeline.py

@@ -2,21 +2,28 @@ from typing import  Dict, List, Any
 from PIL import Image
 import requests
 import torch
+from blip import blip_decoder
 from torchvision import transforms
 from torchvision.transforms.functional import InterpolationMode
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
-from transformers import pipeline, AutoTokenizer
-
 
 class PreTrainedPipeline():
     def __init__(self, path=""):
         # load the optimized model
-        model = ORTModelForSequenceClassification.from_pretrained(path)
-        tokenizer = AutoTokenizer.from_pretrained(path)
-        # create inference pipeline
-        self.pipeline = pipeline("text-classification", model=model, tokenizer=tokenizer)
+        self.model_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_large_caption.pth'
+        self.model = blip_decoder(pretrained=self.model_url, image_size=384, vit='large')
+        self.model.eval()
+        self.model = model.to(device)
+        
+        image_size = 384
+        self.transform = transforms.Compose([
+            transforms.Resize((image_size,image_size),interpolation=InterpolationMode.BICUBIC),
+            transforms.ToTensor(),
+            transforms.Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711))
+            ]) 
+     
 
 
     def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
@@ -29,13 +36,11 @@ class PreTrainedPipeline():
                 - "label": A string representing what the label/class is. There can be multiple labels.
                 - "score": A score between 0 and 1 describing how confident the model is for this label/class.
         """
-        inputs = data.pop("inputs", data)
+        image = data.pop("inputs", data)
         parameters = data.pop("parameters", None)
 
-        # pass inputs with all kwargs in data
-        if parameters is not None:
-            prediction = self.pipeline(inputs, **parameters)
-        else:
-            prediction = self.pipeline(inputs)
+        image = transform(image).unsqueeze(0).to(device)   
+        with torch.no_grad():
+            caption = self.model.generate(image, sample=True, top_p=0.9, max_length=20, min_length=5)
         # postprocess the prediction
-        return prediction
+        return caption