sharding_stage2_pfp16 (#37836)

python/paddle/distributed/fleet/meta_optimizers/dygraph_optimizer/sharding_optimizer_stage2.py

@@ -83,8 +83,14 @@ class ShardingOptimizerStage2(Optimizer):
         # Default information
         self._optim_defaults = kw
         self._optim = optim
+        assert hasattr(self._optim, "_master_weights"
+                       ), "Must use optimizer with _master_weights attribute"
         self._local_params = params
         self._default_device = device
+        self._pfp16 = len(
+            list(
+                filter(lambda x: x.trainable and x.dtype == Type.fp16.value,
+                       self._local_params))) > 0
 
         assert group is not None, "Distributed communication group is must be gived"
         self.group = group
@@ -98,6 +104,12 @@ class ShardingOptimizerStage2(Optimizer):
         # Update optimizer parameters and adjust parameter storage and use according to rank.
         self.update_opt_status()
 
+    def _generate_master_params(self, trainable_params):
+        for param in trainable_params:
+            if param.dtype == Type.fp16.value:
+                self._optim._master_weights[param.name] = paddle.cast(
+                    param, Type.fp32.value)
+
     def update_opt_status(self):
         """Update optimizer status and parameter storage information, and special functions to be developed.
         """
@@ -207,6 +219,8 @@ class ShardingOptimizerStage2(Optimizer):
                     # Merge all the trainable params in a single InternalStorage
                     trainable_params = list(
                         filter(lambda x: x.trainable, params))
+                    if self._pfp16 and dst_rank == self.rank:
+                        self._generate_master_params(trainable_params)
                     if trainable_params:
                         param_storage = ParamStorage(
                             size=self.rank_buffer_size[dtype][dst_rank],

python/paddle/distributed/fleet/meta_parallel/sharding/sharding_stage2.py

@@ -30,6 +30,7 @@ from paddle import nn
 import paddle.distributed as dist
 
 from ...utils.internal_storage import GradStorage
+from ...meta_optimizers.dygraph_optimizer.sharding_optimizer_stage2 import ShardingOptimizerStage2
 from .sharding_utils import Taskflow, Type
 
 
@@ -70,6 +71,11 @@ class ShardingStage2(nn.Layer):
         self._layer = layer
         self._sharding_optimizers = [sharding_optimizer] if not isinstance(
             sharding_optimizer, list) else sharding_optimizer
+        assert all(
+            list(
+                map(lambda opt: isinstance(opt, ShardingOptimizerStage2),
+                    self._sharding_optimizers))
+        ), "Please use ShardingOptimizerStage2 optimizer"
         self._sync_buffers = sync_buffers
         self._auto_refresh_trainable = auto_refresh_trainable
 
@@ -88,8 +94,7 @@ class ShardingStage2(nn.Layer):
 
         # Global statistical parameters
         self._all_params = list(
-            chain(
-                * [optim.local_params for optim in self._sharding_optimizers]))
+            chain(*[optim.local_params for optim in self._sharding_optimizers]))
         self._trainable_params = []
         self._grad_reduced = []
         self._trainable_param2rank = {}
@@ -436,7 +441,7 @@ class ShardingStage2(nn.Layer):
                            ._fill))
 
         self._grad_storage_list = list(
-            chain(* [
+            chain(*[
                 self._grad_storages[dtype].values()
                 for dtype in self._grad_storages.keys()
             ]))

python/paddle/fluid/tests/unittests/dygraph_sharding_stage2.py

@@ -24,7 +24,6 @@ from paddle.fluid.dygraph.nn import Linear
 from paddle.distributed import fleet
 from paddle.fluid.dygraph import nn
 
-from paddle.distributed.fleet.meta_optimizers.dygraph_optimizer.dygraph_sharding_optimizer import DygraphShardingOptimizer
 from paddle.distributed.fleet.meta_optimizers.dygraph_optimizer.sharding_optimizer_stage2 import ShardingOptimizerStage2
 from paddle.distributed.fleet.meta_parallel.sharding.sharding_stage2 import ShardingStage2
 
@@ -70,7 +69,7 @@ def reader_decorator():
     return __reader__
 
 
-def optimizer_setting(model, use_pure_fp16, stage=1):
+def optimizer_setting(model, use_pure_fp16):
     clip = paddle.nn.ClipGradByGlobalNorm(clip_norm=1.0)
     optimizer = paddle.optimizer.AdamW(
         parameters=model.parameters(),
@@ -87,20 +86,16 @@ def train_mlp(model,
               use_pure_fp16=False,
               all_test=False,
               accumulate_grad=False):
-    if sharding_stage == 1:
+    if sharding_stage == "dp":
         hcg = fleet.get_hybrid_communicate_group()
         group = hcg.get_check_parallel_group()
     else:
         group = paddle.distributed.new_group([0, 1])
-    optimizer = optimizer_setting(
-        model=model, use_pure_fp16=use_pure_fp16, stage=sharding_stage)
+    optimizer = optimizer_setting(model=model, use_pure_fp16=use_pure_fp16)
 
     if use_pure_fp16:
-        model, optimizer = paddle.amp.decorate(
-            models=model,
-            optimizers=optimizer,
-            level='O2',
-            save_dtype='float32')
+        model = paddle.amp.decorate(
+            models=model, level='O2', save_dtype='float32')
 
     if sharding_stage == 2:
         optimizer = ShardingOptimizerStage2(
@@ -164,7 +159,7 @@ def train_mlp(model,
     return model.parameters()
 
 
-def test_stage1_stage2():
+def test_dp_stage2():
     mlp = MLP()
     state_dict = mlp.state_dict()
     mlp1 = MLP()
@@ -175,11 +170,13 @@ def test_stage1_stage2():
     mlp2.set_state_dict(state_dict)
     mlp3.set_state_dict(state_dict)
     mlp4.set_state_dict(state_dict)
-    stage1_params = train_mlp(mlp, sharding_stage=1, use_pure_fp16=False)
-    stage2_params = train_mlp(mlp, sharding_stage=2, use_pure_fp16=False)
-    for i in range(len(stage1_params)):
-        np.testing.assert_allclose(
-            stage1_params[i].numpy(), stage2_params[i].numpy(), rtol=1e-6)
+    dp_params = train_mlp(mlp1, sharding_stage="dp", use_pure_fp16=False)
+    stage2_params = train_mlp(mlp2, sharding_stage=2, use_pure_fp16=False)
+    for i in range(len(dp_params)):
+        for j in range(len(stage2_params)):
+            if dp_params[i].name == stage2_params[j].name:
+                np.testing.assert_allclose(
+                    dp_params[i].numpy(), stage2_params[j].numpy(), rtol=1e-6)
 
     stage2_params = train_mlp(
         mlp3, sharding_stage=2, use_pure_fp16=True, all_test=True)
@@ -201,4 +198,4 @@ def test_stage1_stage2():
 
 
 if __name__ == '__main__':
-    test_stage1_stage2()
+    test_dp_stage2()