File size: 5,452 Bytes
243b4bc
 
 
24efe34
243b4bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24efe34
 
 
 
 
243b4bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
138
from tqdm.auto import tqdm
import torch
import os
from warnings import deprecated
from src.models.muti_model import Multimodal

import pandas as pd

class MyModel:
    def __init__(self, config):
        self.config = config
        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

        self.model = Multimodal(config=config).to(self.device)
        self.optimizer = torch.optim.Adam(params=self.model.parameters(), lr=self.config.learning_rate)

        pos_weight_tensor = None
        if config.train_file_path and os.path.exists(config.train_file_path):
            train_df = pd.read_csv(config.train_file_path)
            pos_weight_val = train_df[train_df['label'] == 0].shape[0] / (train_df[train_df['label'] == 1].shape[0] + 1e-5)
            pos_weight_tensor = torch.tensor([pos_weight_val], dtype=torch.float32).to(self.device)
        self.loss_fn = torch.nn.BCEWithLogitsLoss(pos_weight=pos_weight_tensor)

        self.train_loss = []
        self.val_loss = []
        self.best_val_loss = float('inf')
        self.patience_counter = 0

    def train(self, train_data_loader, val_data_loader):
        self.load_model()
        for epoch in range(self.config.epochs):
            self.model.train()
            train_loss = 0.0

            train_pbar = tqdm(train_data_loader, desc=f"Epoch {epoch+1}/{self.config.epochs} [Train]", leave=False)
            for img_feats, text_feats, labels in train_pbar:
                img_feats = img_feats.to(self.device)
                text_feats = text_feats.to(self.device)
                labels = labels.to(self.device).unsqueeze(1)

                self.optimizer.zero_grad()
                outputs = self.model(img_feats, text_feats)
                loss = self.loss_fn(outputs, labels)

                loss.backward()
                self.optimizer.step()

                train_loss += loss.item()
                train_pbar.set_postfix({"batch_loss": f"{loss.item():.4f}"})

            avg_train_loss = train_loss / len(train_data_loader)
            avg_val_loss = self.validate(val_data_loader)

            self.train_loss.append(avg_train_loss)
            self.val_loss.append(avg_val_loss)

            print(f"Epoch [{epoch+1}/{self.config.epochs}] -> Train Loss: {avg_train_loss:.4f} | Val Loss: {avg_val_loss:.4f}")

            if avg_val_loss < self.best_val_loss:
                self.best_val_loss = avg_val_loss
                self.patience_counter = 0
                self.save_model()
            else:
                self.patience_counter += 1
                if self.patience_counter >= self.config.patience:
                    print(f"Early stopping triggered at epoch {epoch+1}")
                    break

    def validate(self, val_data_loader):
        self.model.eval()
        val_loss = 0.0

        val_pbar = tqdm(val_data_loader, desc="Validating", leave=False)
        with torch.no_grad():
            for img_feats, text_feats, labels in val_pbar:
                img_feats = img_feats.to(self.device)
                text_feats = text_feats.to(self.device)
                labels = labels.to(self.device).unsqueeze(1)

                outputs = self.model(img_feats, text_feats)
                loss = self.loss_fn(outputs, labels)

                val_loss += loss.item()
                val_pbar.set_postfix({"val_batch_loss": f"{loss.item():.4f}"})

        return val_loss / len(val_data_loader)

    def predict(self, img_feats, text_feats):
        self.model.eval()
        with torch.no_grad():
            img_feats = img_feats.to(self.device)
            text_feats = text_feats.to(self.device)

            logits = self.model(img_feats, text_feats)
            probs = torch.sigmoid(logits)
        return probs

        
    @deprecated(
        "predict_emb is deprecated. The new dual-index pipeline stores ImageEncoder "
        "and TextEncoder embeddings separately in Pinecone; use those encoders directly "
        "instead of calling predict_emb."
    )
    def predict_emb(self, img_feats, text_feats):
        self.model.eval()

        with torch.no_grad():
            img_feats = img_feats.to(self.device)
            text_feats = text_feats.to(self.device)

            embedding = self.model(
                img_feats,
                text_feats,
                return_embedding=True
            )

        return embedding    

    def save_model(self):
        os.makedirs(self.config.model_dir, exist_ok=True)
        checkpoint_path = os.path.join(self.config.model_dir, self.config.model_name)
        torch.save({
            'model_state_dict': self.model.state_dict(),
            'optimizer_state_dict': self.optimizer.state_dict(),
            'best_val_loss': self.best_val_loss,
            'train_loss': self.train_loss,
            'val_loss': self.val_loss
        }, checkpoint_path)

    def load_model(self):
        checkpoint_path = os.path.join(self.config.model_dir, self.config.model_name)
        if os.path.exists(checkpoint_path):
            checkpoint = torch.load(checkpoint_path, map_location=self.device)
            self.model.load_state_dict(checkpoint['model_state_dict'])
            self.optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
            self.best_val_loss = checkpoint.get('best_val_loss', float('inf'))
            self.train_loss = checkpoint.get('train_loss', [])
            self.val_loss = checkpoint.get('val_loss', [])