File size: 5,471 Bytes
6077218
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import torch
import numpy as np
from tqdm import tqdm
from torch.optim import AdamW
from base import AbstractModel
from transformers.optimization import get_scheduler
from collections import defaultdict, OrderedDict
from utils import get_file_name, get_total_steps
from evaluator import Evaluator


class BaseTrainer(object):
    def __init__(self, config: dict, model: AbstractModel):
        self.config = config
        self.model = model
        self.accelerator = config['accelerator']
        self.evaluator = Evaluator(config)
        self.saved_model_ckpt = os.path.join(
            self.config['ckpt_dir'],
            get_file_name(self.config, suffix='.pth')
        )
        os.makedirs(os.path.dirname(self.saved_model_ckpt), exist_ok=True)
        self.best_metric = 0
        self.best_epoch = 0
        self.count = 0

        self.checkpoints_deque = []

    def train(self, train_dataloader, val_dataloader):
        optimizer = AdamW(
            self.model.parameters(),
            lr=self.config['lr'],
            weight_decay=self.config['weight_decay'],
        )
        total_n_steps = get_total_steps(self.config, train_dataloader)

        self.model, optimizer, train_dataloader, val_dataloader = self.accelerator.prepare(
            self.model, optimizer, train_dataloader, val_dataloader)
        self.config.pop('accelerator')
        self.accelerator.init_trackers(
            project_name="PreferDiff",
            config=self.config
        )
        n_epochs = np.ceil(total_n_steps / (len(train_dataloader) * self.accelerator.num_processes)).astype(int)
        best_epoch = 0
        best_val_score = -1
        for epoch in range(n_epochs):
            # Training
            self.model.train()
            total_loss = 0.0
            train_progress_bar = tqdm(
                train_dataloader,
                total=len(train_dataloader),
                desc=f"Training - [Epoch {epoch + 1}]",
            )
            for batch in train_progress_bar:
                optimizer.zero_grad()
                outputs = self.model(batch)
                loss = outputs['loss']
                self.accelerator.backward(loss)
                optimizer.step()
                # scheduler.step()
                total_loss = total_loss + loss.item()

            self.accelerator.log({"Loss/train_loss": total_loss / len(train_dataloader)}, step=epoch + 1)

            # Evaluation
            if (epoch + 1) % self.config['eval_interval'] == 0:
                all_results = self.evaluate(val_dataloader, split='val')
                if self.accelerator.is_main_process:
                    for key in all_results:
                        self.accelerator.log({f"Val_Metric/{key}": all_results[key]}, step=epoch + 1)
                    print(all_results)

                val_score = all_results[self.config['val_metric']]
                if val_score > best_val_score:
                    best_val_score = val_score
                    best_epoch = epoch + 1
                    if self.accelerator.is_main_process:
                        if self.config['use_ddp']:  # unwrap model for saving
                            unwrapped_model = self.accelerator.unwrap_model(self.model)
                            torch.save(unwrapped_model.state_dict(), self.saved_model_ckpt)
                        else:
                            torch.save(self.model.state_dict(), self.saved_model_ckpt)
                        print(f'[Epoch {epoch + 1}] Saved model checkpoint to {self.saved_model_ckpt}')
                else:
                    print('Patience for {} Times'.format(epoch + 1 - best_epoch))

                if self.config['patience'] is not None and epoch + 1 - best_epoch >= self.config['patience']:
                    print(f'Early stopping at epoch {epoch + 1}')
                    break
        print(f'Best epoch: {best_epoch}, Best val score: {best_val_score}')

    def evaluate(self, dataloader, split='test'):

        self.model.eval()

        all_results = defaultdict(list)
        val_progress_bar = tqdm(
            dataloader,
            total=len(dataloader),
            desc=f"Eval - {split}",
        )
        for batch in val_progress_bar:
            with torch.no_grad():
                batch = {k: v.to(self.accelerator.device) for k, v in batch.items()}
                if self.config['use_ddp']:  # ddp, gather data from all devices for evaluation
                    preds = self.model.module.predict(batch, n_return_sequences=self.evaluator.maxk)
                    all_preds, all_labels = self.accelerator.gather_for_metrics((preds, batch['labels']))
                    results = self.evaluator.calculate_metrics(all_preds, all_labels)
                else:
                    preds = self.model.predict(batch, n_return_sequences=self.evaluator.maxk)
                    results = self.evaluator.calculate_metrics(preds, batch['labels'])

                for key, value in results.items():
                    all_results[key].append(value)


        output_results = OrderedDict()
        for metric in self.config['metrics']:
            for k in self.config['topk']:
                key = f"{metric}@{k}"
                output_results[key] = torch.cat(all_results[key]).mean().item()
        return output_results

    def end(self):
        """
        Ends the training process and releases any used resources
        """
        self.accelerator.end_training()