Add Linear warmup+decay lr schedule (#414)

Update lr schedule unit tests

deepspeed/runtime/lr_schedules.py

@@ -19,7 +19,8 @@ LR_SCHEDULE = 'lr_schedule'
 LR_RANGE_TEST = 'LRRangeTest'
 ONE_CYCLE = 'OneCycle'
 WARMUP_LR = 'WarmupLR'
-VALID_LR_SCHEDULES = [LR_RANGE_TEST, ONE_CYCLE, WARMUP_LR]
+WARMUP_DECAY_LR = 'WarmupDecayLR'
+VALID_LR_SCHEDULES = [LR_RANGE_TEST, ONE_CYCLE, WARMUP_LR, WARMUP_DECAY_LR]
 
 LR_RANGE_TEST_MIN_LR = 'lr_range_test_min_lr'
 LR_RANGE_TEST_STEP_RATE = 'lr_range_test_step_rate'
@@ -47,6 +48,8 @@ WARMUP_MIN_LR = 'warmup_min_lr'
 WARMUP_MAX_LR = 'warmup_max_lr'
 WARMUP_NUM_STEPS = 'warmup_num_steps'
 
+TOTAL_NUM_STEPS = 'total_num_steps'
+
 
 def add_tuning_arguments(parser):
     group = parser.add_argument_group('Convergence Tuning',
@@ -714,3 +717,54 @@ class WarmupLR(object):
                     FileNotFoundError(param_value)))
             return list(param_value)
         return [param_value] * len(optimizer.param_groups)
+
+
+class WarmupDecayLR(WarmupLR):
+    """Increase the learning rate of each parameter group from min lr to max lr
+        over warmup_num_steps steps, and then decay at linear rate over the remaining training steps.
+
+        Args:
+            optimizer (Optimizer): Wrapped optimizer.
+            total_num_steps (int): total number of training steps
+            warmup_min_lr (float or list): minimum learning rate. Default: 0
+            warmup_max_lr (float or list): maximum learning rate. Default: 0.001
+            warmup_num_steps (int): number of steps to warm up from min_lr to max_lr. Default: 1000
+            last_batch_iteration (int): The index of the last batch. Default: -1.
+        Example:
+            >>> optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)
+            >>> scheduler = WarmupDecayLR(optimizer, 1000000)
+            >>> data_loader = torch.utils.data.DataLoader(...)
+            >>> for epoch in range(10):
+            >>>     for batch in data_loader:
+            >>>         train_batch(...)
+            >>>         scheduler.step()
+
+    """
+    def __init__(self,
+                 optimizer: Optimizer,
+                 total_num_steps: int,
+                 warmup_min_lr: float = 0.0,
+                 warmup_max_lr: float = 0.001,
+                 warmup_num_steps: int = 1000,
+                 last_batch_iteration: int = -1):
+
+        self.total_num_steps = total_num_steps
+        super(WarmupDecayLR,
+              self).__init__(optimizer,
+                             warmup_min_lr,
+                             warmup_max_lr,
+                             warmup_num_steps,
+                             last_batch_iteration)
+        if self.total_num_steps < self.warmup_num_steps:
+            logger.warning('total_num_steps {} is less than warmup_num_steps {}'.format(
+                total_num_steps,
+                warmup_num_steps))
+
+    def _get_gamma(self):
+        if self.last_batch_iteration < self.warmup_num_steps:
+            return self.inverse_log_warm_up * math.log(self.last_batch_iteration + 1)
+        return max(
+            0.0,
+            float(self.total_num_steps - self.last_batch_iteration) /
+            float(max(1.0,
+                      self.total_num_steps - self.warmup_num_steps)))

tests/unit/test_lr_schedulers.py

@@ -6,17 +6,25 @@ import json
 import os
 from common import distributed_test
 from simple_model import SimpleModel, SimpleOptimizer, random_dataloader, args_from_dict
+from deepspeed.runtime.lr_schedules import LR_RANGE_TEST, ONE_CYCLE, WARMUP_LR, WARMUP_DECAY_LR
+from deepspeed.runtime.lr_schedules import WARMUP_MIN_LR, WARMUP_MAX_LR, WARMUP_NUM_STEPS, TOTAL_NUM_STEPS
+from deepspeed.runtime.lr_schedules import CYCLE_MIN_LR, CYCLE_MAX_LR
 
 
 @pytest.mark.parametrize("scheduler_type,params",
-                         [("WarmupLR",
+                         [(WARMUP_LR,
                            {}),
-                          ("OneCycle",
+                          (WARMUP_DECAY_LR,
                            {
-                               'cycle_min_lr': 0,
-                               'cycle_max_lr': 0
+                               WARMUP_NUM_STEPS: 10,
+                               TOTAL_NUM_STEPS: 20
                            }),
-                          ("LRRangeTest",
+                          (ONE_CYCLE,
+                           {
+                               CYCLE_MIN_LR: 0,
+                               CYCLE_MAX_LR: 0
+                           }),
+                          (LR_RANGE_TEST,
                            {})])
 def test_get_lr_before_train(tmpdir, scheduler_type, params):
     config_dict = {
@@ -42,8 +50,8 @@ def test_get_lr_before_train(tmpdir, scheduler_type, params):
     @distributed_test(world_size=[1])
     def _test_get_lr_before_train(args, model, hidden_dim):
         model, _, _, lr_scheduler = deepspeed.initialize(args=args,
-                                             model=model,
-                                             model_parameters=model.parameters())
+                                                         model=model,
+                                                         model_parameters=model.parameters())
         data_loader = random_dataloader(model=model,
                                         total_samples=50,
                                         hidden_dim=hidden_dim,
@@ -57,3 +65,143 @@ def test_get_lr_before_train(tmpdir, scheduler_type, params):
             model.step()
 
     _test_get_lr_before_train(args=args, model=model, hidden_dim=hidden_dim)
+
+
+@pytest.mark.parametrize("warmup_num_steps", [10, 15, 19, 33])
+def test_lr_warmup_schedule(tmpdir, warmup_num_steps):
+    config_dict = {
+        "train_batch_size": 2,
+        "steps_per_print": 1,
+        "optimizer": {
+            "type": "Adam",
+            "params": {
+                "lr": 0.00015
+            },
+        },
+        "scheduler": {
+            "type": WARMUP_LR,
+            "params": {
+                WARMUP_MIN_LR: 0.1,
+                WARMUP_MAX_LR: 0.2,
+                WARMUP_NUM_STEPS: warmup_num_steps
+            }
+        },
+        "gradient_clipping": 1.0
+    }
+
+    total_num_steps = 2 * warmup_num_steps
+
+    args = args_from_dict(tmpdir, config_dict)
+    hidden_dim = 10
+
+    model = SimpleModel(hidden_dim, empty_grad=False)
+
+    @distributed_test(world_size=[1])
+    def _test_lr_warmup_schedule(args, model, hidden_dim, schedule_params, num_steps):
+        model, _, _, lr_scheduler = deepspeed.initialize(args=args,
+                                                         model=model,
+                                                         model_parameters=model.parameters())
+
+        data_loader = random_dataloader(model=model,
+                                        total_samples=num_steps * 2,
+                                        hidden_dim=hidden_dim,
+                                        device=model.device,
+                                        dtype=torch.float)
+        step_lrs = []
+        for n, batch in enumerate(data_loader):
+            loss = model(batch[0], batch[1])
+            model.backward(loss)
+            model.step()
+            step_lrs.append(lr_scheduler.get_lr())
+
+        # Verify initial lr
+        assert step_lrs[0] == [schedule_params[WARMUP_MIN_LR]]
+
+        # Verify warmup completion
+        warmup_num_steps = schedule_params[WARMUP_NUM_STEPS]
+        warmup_max_lr = [schedule_params[WARMUP_MAX_LR]]
+        assert step_lrs[warmup_num_steps] == warmup_max_lr
+
+        # Verify post-warmup completion
+        assert all([warmup_max_lr == lr for lr in step_lrs[warmup_num_steps:]])
+
+    _test_lr_warmup_schedule(args=args,
+                             model=model,
+                             hidden_dim=hidden_dim,
+                             schedule_params=config_dict["scheduler"]["params"],
+                             num_steps=total_num_steps)
+
+
+@pytest.mark.parametrize("warmup_num_steps", [10, 15, 19, 33])
+def test_lr_warmup_decay_schedule(tmpdir, warmup_num_steps):
+    config_dict = {
+        "train_batch_size": 2,
+        "steps_per_print": 1,
+        "optimizer": {
+            "type": "Adam",
+            "params": {
+                "lr": 0.00015
+            },
+        },
+        "scheduler": {
+            "type": WARMUP_DECAY_LR,
+            "params": {
+                WARMUP_MIN_LR: 0.1,
+                WARMUP_MAX_LR: 0.2,
+                WARMUP_NUM_STEPS: warmup_num_steps,
+                TOTAL_NUM_STEPS: warmup_num_steps * 2
+            }
+        },
+        "gradient_clipping": 1.0
+    }
+
+    args = args_from_dict(tmpdir, config_dict)
+    hidden_dim = 10
+
+    model = SimpleModel(hidden_dim, empty_grad=False)
+
+    @distributed_test(world_size=[1])
+    def _test_lr_warmup_decay_schedule(args,
+                                       model,
+                                       hidden_dim,
+                                       schedule_params,
+                                       num_steps):
+        model, _, _, lr_scheduler = deepspeed.initialize(args=args,
+                                                         model=model,
+                                                         model_parameters=model.parameters())
+
+        data_loader = random_dataloader(model=model,
+                                        total_samples=num_steps * 2,
+                                        hidden_dim=hidden_dim,
+                                        device=model.device,
+                                        dtype=torch.float)
+        step_lrs = []
+        for n, batch in enumerate(data_loader):
+            loss = model(batch[0], batch[1])
+            model.backward(loss)
+            model.step()
+            step_lrs.append(lr_scheduler.get_lr())
+
+        # Verify initial lr
+        assert step_lrs[0] == [schedule_params[WARMUP_MIN_LR]]
+
+        # Verify lr at warmup completion
+        warmup_num_steps = schedule_params[WARMUP_NUM_STEPS]
+        warmup_max_lr = [schedule_params[WARMUP_MAX_LR]]
+        assert step_lrs[warmup_num_steps] == warmup_max_lr
+
+        # Verify decay phase
+        previous_lr = warmup_max_lr
+        for lr in step_lrs[warmup_num_steps + 1:]:
+            assert lr < previous_lr
+            previous_lr = lr
+
+    schedule_params = config_dict["scheduler"]["params"]
+
+    total_num_steps = schedule_params[TOTAL_NUM_STEPS]
+
+    _test_lr_warmup_decay_schedule(args=args,
+                                   model=model,
+                                   hidden_dim=hidden_dim,
+                                   schedule_params=schedule_params,
+                                   num_steps=total_num_steps)