super-gradients/tests/unit_tests/lr_warmup_test.py

import unittest

import numpy as np

from super_gradients.training import SgModel
from super_gradients.training.metrics import Accuracy
from super_gradients.training.datasets import ClassificationTestDatasetInterface
from super_gradients.training.models import LeNet
from super_gradients.training.utils.callbacks import TestLRCallback, LRCallbackBase, Phase


class ExponentialWarmupLRCallback(LRCallbackBase):
    """
    LR scheduling callback for exponential warmup.
    LR grows exponentially from warmup_initial_lr to initial lr.
    When warmup_initial_lr is None- LR climb starts from 0.001
    """

    def __init__(self, **kwargs):
        super().__init__(Phase.TRAIN_EPOCH_START, **kwargs)
        self.warmup_initial_lr = self.training_params.warmup_initial_lr or 0.001
        warmup_epochs = self.training_params.lr_warmup_epochs
        lr_start = self.warmup_initial_lr
        lr_end = self.initial_lr
        self.c1 = (lr_end - lr_start) / (np.exp(warmup_epochs) - 1.)
        self.c2 = (lr_start * np.exp(warmup_epochs) - lr_end) / (np.exp(warmup_epochs) - 1.)

    def perform_scheduling(self, context):
        self.lr = self.c1 * np.exp(context.epoch) + self.c2
        self.update_lr(context.optimizer, context.epoch, None)

    def is_lr_scheduling_enabled(self, context):
        return self.training_params.lr_warmup_epochs >= context.epoch


class LRWarmupTest(unittest.TestCase):
    def setUp(self) -> None:
        self.dataset_params = {"batch_size": 4}
        self.dataset = ClassificationTestDatasetInterface(dataset_params=self.dataset_params)
        self.arch_params = {'num_classes': 10}

    def test_lr_warmup(self):
        # Define Model
        net = LeNet()
        model = SgModel("lr_warmup_test", model_checkpoints_location='local')
        model.connect_dataset_interface(self.dataset)
        model.build_model(net, arch_params=self.arch_params)

        lrs = []
        phase_callbacks = [TestLRCallback(lr_placeholder=lrs)]

        train_params = {"max_epochs": 5, "lr_updates": [], "lr_decay_factor": 0.1, "lr_mode": "step",
                        "lr_warmup_epochs": 3, "initial_lr": 1, "loss": "cross_entropy", "optimizer": 'SGD',
                        "criterion_params": {}, "optimizer_params": {"weight_decay": 1e-4, "momentum": 0.9},
                        "train_metrics_list": [Accuracy()], "valid_metrics_list": [Accuracy()],
                        "loss_logging_items_names": ["Loss"], "metric_to_watch": "Accuracy",
                        "greater_metric_to_watch_is_better": True, "ema": False, "phase_callbacks": phase_callbacks,
                        "warmup_mode": "linear_step"}

        expected_lrs = [0.25, 0.5, 0.75, 1.0, 1.0]
        model.train(train_params)
        self.assertListEqual(lrs, expected_lrs)

    def test_lr_warmup_with_lr_scheduling(self):
        # Define Model
        net = LeNet()
        model = SgModel("lr_warmup_test", model_checkpoints_location='local')
        model.connect_dataset_interface(self.dataset)
        model.build_model(net, arch_params=self.arch_params)

        lrs = []
        phase_callbacks = [TestLRCallback(lr_placeholder=lrs)]

        train_params = {"max_epochs": 5, "cosine_final_lr_ratio": 0.2, "lr_mode": "cosine",
                        "lr_warmup_epochs": 3, "initial_lr": 1, "loss": "cross_entropy", "optimizer": 'SGD',
                        "criterion_params": {}, "optimizer_params": {"weight_decay": 1e-4, "momentum": 0.9},
                        "train_metrics_list": [Accuracy()], "valid_metrics_list": [Accuracy()],
                        "loss_logging_items_names": ["Loss"], "metric_to_watch": "Accuracy",
                        "greater_metric_to_watch_is_better": True, "ema": False, "phase_callbacks": phase_callbacks,
                        "warmup_mode": "linear_step"}

        expected_lrs = [0.25, 0.5, 0.75, 0.9236067977499791, 0.4763932022500211]
        model.train(train_params)

        # ALTHOUGH NOT SEEN IN HERE, THE 4TH EPOCH USES LR=1, SO THIS IS THE EXPECTED LIST AS WE COLLECT
        # THE LRS AFTER THE UPDATE
        self.assertListEqual(lrs, expected_lrs)

    def test_warmup_initial_lr(self):
        # Define Model
        net = LeNet()
        model = SgModel("test_warmup_initial_lr", model_checkpoints_location='local')
        model.connect_dataset_interface(self.dataset)
        model.build_model(net, arch_params=self.arch_params)

        lrs = []
        phase_callbacks = [TestLRCallback(lr_placeholder=lrs)]

        train_params = {"max_epochs": 5, "lr_updates": [], "lr_decay_factor": 0.1, "lr_mode": "step",
                        "lr_warmup_epochs": 3, "loss": "cross_entropy", "optimizer": 'SGD',
                        "criterion_params": {}, "optimizer_params": {"weight_decay": 1e-4, "momentum": 0.9},
                        "train_metrics_list": [Accuracy()], "valid_metrics_list": [Accuracy()],
                        "loss_logging_items_names": ["Loss"], "metric_to_watch": "Accuracy",
                        "greater_metric_to_watch_is_better": True, "ema": False, "phase_callbacks": phase_callbacks,
                        "warmup_mode": "linear_step", "initial_lr": 1, "warmup_initial_lr": 4.}

        expected_lrs = [4., 3., 2., 1., 1.]
        model.train(train_params)
        self.assertListEqual(lrs, expected_lrs)

    def test_custom_lr_warmup(self):
        # Define Model
        net = LeNet()
        model = SgModel("custom_lr_warmup_test", model_checkpoints_location='local')
        model.connect_dataset_interface(self.dataset)
        model.build_model(net, arch_params=self.arch_params)

        lrs = []
        phase_callbacks = [TestLRCallback(lr_placeholder=lrs)]

        train_params = {"max_epochs": 5, "lr_updates": [], "lr_decay_factor": 0.1, "lr_mode": "step",
                        "lr_warmup_epochs": 3, "loss": "cross_entropy", "optimizer": 'SGD',
                        "criterion_params": {}, "optimizer_params": {"weight_decay": 1e-4, "momentum": 0.9},
                        "train_metrics_list": [Accuracy()], "valid_metrics_list": [Accuracy()],
                        "loss_logging_items_names": ["Loss"], "metric_to_watch": "Accuracy",
                        "greater_metric_to_watch_is_better": True, "ema": False, "phase_callbacks": phase_callbacks,
                        "warmup_mode": ExponentialWarmupLRCallback, "initial_lr": 1., "warmup_initial_lr": 0.1}

        expected_lrs = [0.1, 0.18102751585334242, 0.40128313980266034, 1.0, 1.0]
        model.train(train_params)
        self.assertListEqual(lrs, expected_lrs)


if __name__ == '__main__':
    unittest.main()