File size: 4,759 Bytes
7569ad7
 
 
 
 
7755c1c
7569ad7
7755c1c
7569ad7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7755c1c
7569ad7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
import os
from pprint import pprint
from configs.config import parser
from dataset.data_module import DataModule
from models.R2GenGPT import R2GenGPT
import torch
from transformers import BertTokenizer, AutoImageProcessor
from PIL import Image
import numpy as np
import streamlit as st
from lightning.pytorch import seed_everything

# Initialize the app
# st.title("Chest X-ray Report Generator")

# Function to load the model
def load_model(args):
    model = R2GenGPT(args)
    model.eval()
    model.freeze()
    return model

# Function to parse image
def _parse_image(vit_feature_extractor, img):
    pixel_values = vit_feature_extractor(img, return_tensors="pt").pixel_values
    return pixel_values[0]

# Function to generate predictions
def generate_predictions(image_path, vit_feature_extractor, model):
    model.llama_tokenizer.padding_side = "right"

    with Image.open(image_path) as pil:
        array = np.array(pil, dtype=np.uint8)
        if array.shape[-1] != 3 or len(array.shape) != 3:
            array = np.array(pil.convert("RGB"), dtype=np.uint8)
        image = _parse_image(vit_feature_extractor, array)
        image = image.unsqueeze(0) 
        # image = image[None, :]
        image = image.to(device='cuda:0')

    print("Model Encoding for Image: ", model.encode_img(image))
    try:
        img_embeds, atts_img = model.encode_img(image)
        print("Image embeddings in try blk", img_embeds)

        print("Try block for Image Embeddings \n")
        
    except Exception as e:
        st.error(e)
        print(st.error(e))
        print("Except block for Image embeddings \n")
       # return []

    img_embeds = model.layer_norm(img_embeds)
    img_embeds, atts_img = model.prompt_wrap(img_embeds, atts_img)
    print("Image embeddings: ", img_embeds)


    batch_size = img_embeds.shape[0]
    print("Batch size printed: ", batch_size) 
    bos = torch.ones([batch_size, 1],
                     dtype=atts_img.dtype,
                     device=atts_img.device) * model.llama_tokenizer.bos_token_id
    bos_embeds = model.embed_tokens(bos)
    atts_bos = atts_img[:, :1]
    print("Attention: ", atts_bos)

    inputs_embeds = torch.cat([bos_embeds, img_embeds], dim=1)
    print("Shape of Input emb", inputs_embeds)
    inputs_embeds = inputs_embeds.type(torch.float16)
    attention_mask = torch.cat([atts_bos, atts_img], dim=1)
    print("Shape of Attention mask: ", attention_mask)

    try:
        with torch.no_grad():
            outputs = model.llama_model.generate(inputs_embeds=inputs_embeds)
            print("output", outputs)
    except Exception as e:
        st.error(e)
        return []

    hypo = [model.decode(i) for i in outputs]
    print("Generated Report :", hypo)
    return hypo

# Function to perform inference
def inference(args, uploaded_file):
    model = load_model(args)
    vit_feature_extractor = AutoImageProcessor.from_pretrained(args.vision_model)
    
    with open("/workspace/p10_p10046166_s50051329_427446c1-881f5cce-85191ce1-91a58ba9-0a57d3f5.jpg", "wb") as f:
        f.write(uploaded_file.getbuffer())

    predictions = generate_predictions("/workspace/p10_p10046166_s50051329_427446c1-881f5cce-85191ce1-91a58ba9-0a57d3f5.jpg", vit_feature_extractor, model)
    print("Predictions: ", predictions)
    os.remove("/workspace/p10_p10046166_s50051329_427446c1-881f5cce-85191ce1-91a58ba9-0a57d3f5.jpg")

    return predictions

# Main function
def main():
    #parser = argparse.ArgumentParser()
    # other arguments
    #parser.add_argument('--file', type=open, action=LoadFromFile)

    args = parser.parse_args()
    pprint(vars(args))
    seed_everything(42, workers=True)

    # File uploader for image
    model = load_model(args)
    vit_feature_extractor = AutoImageProcessor.from_pretrained(args.vision_model)
    predictions = generate_predictions("/workspace/p10_p10046166_s57379357_6e511483-c7e1601c-76890b2f-b0c6b55d-e53bcbf6.jpg", vit_feature_extractor, model)
    print("Predictions: ", predictions)

    print("Inference: ", inference(args,  "/workspace/p10_p10046166_s57379357_6e511483-c7e1601c-76890b2f-b0c6b55d-e53bcbf6.jpg"))
    #uploaded_file = st.file_uploader("Choose a chest X-ray image...", type="jpg")

    #if uploaded_file is not None:
    #    st.image(uploaded_file, caption='Uploaded Image.', use_column_width=True)
    #    st.write("")
    #    st.write("Generating report...")

        #predictions = inference(args, uploaded_file)

     #   if predictions:
     #       st.write("Generated Report:")
           
     #       for pred in predictions:
     #          print("Generated Report", pred)
    #          st.write(pred)
    #    else:
    #        st.write("Failed to generate report.")

if __name__ == '__main__':
    main()