[fix]: handle inputs with containers in mixed precision (#486)

* [fix]: handle inputs with containers

- this is an issue surfaces by vissl as well
- fix seems to be super simple
- also cleaned up two tests with respect to multiple such tests
  running back to back (they don't do that presently)

* cleanup

* fix

* lint

fairscale/nn/data_parallel/fully_sharded_data_parallel.py

@@ -20,13 +20,7 @@ import torch.nn.functional as F
 
 from fairscale.nn.misc import FlattenParamsWrapper
 from fairscale.optim.utils import calc_grad_norm
-from fairscale.utils.containers import (
-    apply_to_tensors,
-    pack_kwargs,
-    split_non_tensors,
-    unpack_kwargs,
-    unpack_non_tensors,
-)
+from fairscale.utils.containers import apply_to_tensors
 from fairscale.utils.parallel import chunk_and_pad, validate_process_group
 from fairscale.utils.reduce_scatter_bucketer import ReduceScatterBucketer
 from fairscale.utils.state_dict import replace_by_prefix_
@@ -1189,15 +1183,14 @@ class FullyShardedDataParallel(nn.Module):
 def cast_inputs_to_fp16(*args: Any, **kwargs: Any) -> Tuple[Any, Any]:
     """
     Cast any Tensors in *args or **kwargs to FP16.
-
-    Doesn't currently support Tensors nested inside containers (e.g., dict).
     """
-    kwarg_keys, flat_args = pack_kwargs(*args, **kwargs)
-    tensor_inputs, packed_non_tensor_inputs = split_non_tensors(flat_args)
-    tensor_inputs = tuple(t.half() if torch.is_floating_point(t) else t for t in tensor_inputs)
-    flat_args = unpack_non_tensors(tensor_inputs, packed_non_tensor_inputs)
-    args, kwargs = unpack_kwargs(kwarg_keys, flat_args)
-    return args, kwargs
+
+    def fn(x: torch.Tensor) -> torch.Tensor:
+        if x.dtype is torch.float32:
+            return x.half()
+        return x
+
+    return apply_to_tensors(fn, args), apply_to_tensors(fn, kwargs)
 
 
 def cast_buffers_(

tests/ci_test_list_3.txt

@@ -2,6 +2,7 @@ tests/nn/data_parallel/test_fsdp_uneven.py
 tests/nn/data_parallel/test_fsdp_grad_scaler.py
 tests/nn/data_parallel/test_fsdp_no_sync.py
 tests/nn/data_parallel/test_fsdp_summon_full_params.py
+tests/nn/data_parallel/test_fsdp_input.py
 tests/nn/data_parallel/test_sharded_ddp_features.py
 tests/nn/data_parallel/test_sharded_ddp_pytorch_parity.py
 tests/nn/pipe/skip/test_gpipe.py

tests/nn/data_parallel/test_fsdp_grad_scaler.py

+# Copyright (c) Facebook, Inc. and its affiliates. All rights reserved.
+#
+# This source code is licensed under the BSD license found in the
+# LICENSE file in the root directory of this source tree.
+
+# pylint: disable=missing-module-docstring
+# pylint: disable=missing-class-docstring
+# pylint: disable=missing-function-docstring
+
+""" Test FSDP with grad scaler. """
+
 import os
-from unittest import mock
+import random
 
 import pytest
 import torch
@@ -17,15 +28,18 @@ except ImportError:
     pytestmark = pytest.mark.skip
 
 
-@mock.patch.dict(os.environ, {"MASTER_ADDR": "localhost", "MASTER_PORT": "1337"}, clear=True)
+# Mixed precision needs cuda.
 @skip_if_no_cuda
 def test_scaler_cpu_offload_breaks():
-
     device = torch.device("cuda")
     torch.cuda.set_device(0)
 
+    # Random port in case the next test run quickly, same port would cause conflict.
+    os.environ["MASTER_ADDR"] = "localhost"
+    os.environ["MASTER_PORT"] = str(random.randint(2000, 3000))
     torch.distributed.init_process_group(backend="nccl", rank=0, world_size=1)
 
+    try:
         scaler = ShardedGradScaler()
         model = FullyShardedDataParallel(nn.Linear(5, 5), cpu_offload=True, mixed_precision=True)
         optim = torch.optim.SGD(model.parameters(), lr=1e-3)
@@ -41,4 +55,9 @@ def test_scaler_cpu_offload_breaks():
         with pytest.raises(RuntimeError):
             scaler.step(optim)
             scaler.update()
+
+    finally:
+        # Clean-up is important or the next test in this file may fail to init the PG.
         torch.distributed.destroy_process_group()
+        del os.environ["MASTER_ADDR"]
+        del os.environ["MASTER_PORT"]

tests/nn/data_parallel/test_fsdp_input.py

+# Copyright (c) Facebook, Inc. and its affiliates. All rights reserved.
+#
+# This source code is licensed under the BSD license found in the
+# LICENSE file in the root directory of this source tree.
+
+# pylint: disable=missing-module-docstring
+# pylint: disable=missing-class-docstring
+# pylint: disable=missing-function-docstring
+
+""" Test FSDP with different input types. """
+
+import os
+import random
+
+import pytest
+import torch
+from torch.nn import Linear, Module
+from torch.optim import SGD
+
+from fairscale.nn.data_parallel import FullyShardedDataParallel as FSDP
+from fairscale.nn.data_parallel import TrainingState
+from fairscale.utils.testing import skip_if_no_cuda, torch_version
+
+
+# We only test on GPU since mix-precision only really works on GPU.
+@skip_if_no_cuda
+@pytest.mark.parametrize(
+    "fsdp_config", [{}, {"mixed_precision": True}],
+)
+@pytest.mark.parametrize("input_cls", [dict, list])
+def test_it(fsdp_config, input_cls):
+    """Test FSDP with input being a list or a dict, only single GPU."""
+    if torch_version() < (1, 6, 0):
+        pytest.skip("older pytorch doesn't support reduce_scatter")
+
+    # Random port in case the next test run quickly, same port would cause conflict.
+    os.environ["MASTER_ADDR"] = "localhost"
+    os.environ["MASTER_PORT"] = str(random.randint(2000, 3000))
+    torch.distributed.init_process_group(backend="nccl", rank=0, world_size=1)
+
+    try:
+        assert isinstance(fsdp_config, dict), str(fsdp_config)
+
+        class Model(Module):
+            def __init__(self):
+                super().__init__()
+                self.layer = Linear(4, 4)
+
+            def forward(self, input):
+                if isinstance(input, list):
+                    input = input[0]
+                else:
+                    assert isinstance(input, dict), input
+                    input = input["in"]
+                return self.layer(input)
+
+        model = FSDP(Model(), **fsdp_config).cuda()
+        optim = SGD(model.parameters(), lr=0.1)
+
+        for _ in range(5):
+            in_data = torch.rand(64, 4).cuda()
+            in_data.requires_grad = True
+            if input_cls is list:
+                in_data = [in_data]
+            else:
+                assert input_cls is dict
+                in_data = {"in": in_data}
+
+            out = model(in_data)
+            out.sum().backward()
+            optim.step()
+            optim.zero_grad()
+
+        model.assert_state(TrainingState.IDLE)
+
+    finally:
+        # Clean-up is important or the next test in this file may fail to init the PG.
+        torch.distributed.destroy_process_group()
+        del os.environ["MASTER_ADDR"]
+        del os.environ["MASTER_PORT"]

tests/nn/wrap/test_wrap.py

@@ -5,8 +5,8 @@
 
 import functools
 import os
+import random
 import unittest
-from unittest import mock
 
 import torch
 import torch.nn as nn
@@ -16,13 +16,14 @@ from fairscale.nn import FullyShardedDataParallel as FSDP
 from fairscale.nn import auto_wrap, default_auto_wrap_policy, enable_wrap, wrap
 from fairscale.utils.testing import DummyProcessGroup
 
+try:
+    from torch.cuda.amp import autocast
+except ImportError:
+    autocast = None  # type: ignore
+
 
 class TestAutoWrap(unittest.TestCase):
     def setUp(self) -> None:
-        version = torch.__version__.split(".")[:2]
-        major, minor = int(version[0]), int(version[1])
-        if major < 1 or (major == 1 and minor < 6):
-            raise unittest.SkipTest("Need pytorch version >= 1.6 due to autocast")
         self.process_group = DummyProcessGroup(rank=0, size=1)
 
     def test_wrap(self):
@@ -129,15 +130,18 @@ class TestAutoWrap(unittest.TestCase):
         """
         self._auto_wrap_smoke_test(enable_mixed_precision=True)
 
-    @mock.patch.dict(os.environ, {"MASTER_ADDR": "localhost", "MASTER_PORT": "12345"}, clear=True)
     @unittest.skipIf(not torch.cuda.is_available(), "Test Requires CUDA")
+    @unittest.skipIf(autocast is None, "Test Requires autocast")
     def _auto_wrap_smoke_test(self, enable_mixed_precision):
-        from torch.cuda.amp import autocast
-
         device = torch.device("cuda")
         torch.cuda.set_device(0)
+
+        # Random port in case the next test run quickly, same port would cause conflict.
+        os.environ["MASTER_ADDR"] = "localhost"
+        os.environ["MASTER_PORT"] = str(random.randint(2000, 3000))
         torch.distributed.init_process_group(backend="nccl", rank=0, world_size=1)
 
+        try:
             with enable_wrap(wrapper_cls=FSDP, mixed_precision=enable_mixed_precision):
                 sequential = nn.Sequential(
                     nn.Linear(5, 5), nn.Linear(5, 5), nn.Sequential(nn.Linear(5, 5), nn.Linear(5, 5))
@@ -151,4 +155,7 @@ class TestAutoWrap(unittest.TestCase):
                 output = model(input)
                 loss = F.mse_loss(input, output)
             loss.backward()
+        finally:
             torch.distributed.destroy_process_group()
+            del os.environ["MASTER_ADDR"]
+            del os.environ["MASTER_PORT"]