Skip to content

Re-land torchrec pipeline refactor (with fix) #2853

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Re-land torchrec pipeline refactor (with fix) #2853

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions torchrec/distributed/train_pipeline/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,8 @@
_to_device, # noqa
_wait_for_batch, # noqa
ArgInfo, # noqa
ArgInfoStepFactory, # noqa
CallArgs, # noqa
DataLoadingThread, # noqa
In, # noqa
Out, # noqa
Expand Down
150 changes: 80 additions & 70 deletions torchrec/distributed/train_pipeline/tests/test_train_pipelines.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,9 +63,13 @@
DataLoadingThread,
EmbeddingPipelinedForward,
get_h2d_func,
GetAttrArgInfoStep,
GetItemArgInfoStep,
NoopArgInfoStep,
PipelinedForward,
PipelinedPostproc,
PipelineStage,
PostprocArgInfoStep,
SparseDataDistUtil,
StageOut,
TrainPipelineContext,
Expand Down Expand Up @@ -1152,44 +1156,56 @@ def test_pipeline_postproc_not_shared_with_arg_transform(self) -> None:
pipelined_weighted_ebc = pipeline._pipelined_modules[1]

# Check pipelined args
for ebc in [pipelined_ebc, pipelined_weighted_ebc]:
self.assertEqual(len(ebc.forward._args), 1)
self.assertEqual(ebc.forward._args[0].input_attrs, ["", 0])
self.assertEqual(ebc.forward._args[0].is_getitems, [False, True])
self.assertEqual(len(ebc.forward._args[0].postproc_modules), 2)
self.assertIsInstance(
ebc.forward._args[0].postproc_modules[0], PipelinedPostproc
)
self.assertEqual(ebc.forward._args[0].postproc_modules[1], None)

self.assertEqual(len(pipelined_ebc.forward._args.args), 1)
self.assertEqual(len(pipelined_ebc.forward._args.kwargs), 0)
self.assertEqual(
pipelined_ebc.forward._args[0].postproc_modules[0],
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_nonweighted`.
pipelined_model.module.postproc_nonweighted,
pipelined_ebc.forward._args.args[0].steps,
[
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `_postproc_module`.
PostprocArgInfoStep(pipelined_model.module.postproc_nonweighted),
GetItemArgInfoStep(0),
],
)
self.assertEqual(len(pipelined_weighted_ebc.forward._args.args), 1)
self.assertEqual(len(pipelined_weighted_ebc.forward._args.kwargs), 0)
self.assertEqual(
pipelined_weighted_ebc.forward._args[0].postproc_modules[0],
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_weighted`.
pipelined_model.module.postproc_weighted,
pipelined_weighted_ebc.forward._args.args[0].steps,
[
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `_postproc_module`.
PostprocArgInfoStep(pipelined_model.module.postproc_weighted),
GetItemArgInfoStep(0),
],
)

# postproc args
self.assertEqual(len(pipeline._pipelined_postprocs), 2)
input_attr_names = {"idlist_features", "idscore_features"}
for i in range(len(pipeline._pipelined_postprocs)):
postproc_mod = pipeline._pipelined_postprocs[i]
self.assertEqual(len(postproc_mod._args), 1)
# postprocs can be added in any order, so we can't assert on exact steps structures
self.assertEqual(len(pipeline._pipelined_postprocs[0]._args.args), 1)
self.assertEqual(len(pipeline._pipelined_postprocs[0]._args.kwargs), 0)
self.assertEqual(len(pipeline._pipelined_postprocs[0]._args.args[0].steps), 2)
self.assertEqual(
pipeline._pipelined_postprocs[0]._args.args[0].steps[0], NoopArgInfoStep()
)
self.assertIsInstance(
pipeline._pipelined_postprocs[0]._args.args[0].steps[1], GetAttrArgInfoStep
)

input_attr_name = postproc_mod._args[0].input_attrs[1]
self.assertTrue(input_attr_name in input_attr_names)
self.assertEqual(postproc_mod._args[0].input_attrs, ["", input_attr_name])
input_attr_names.remove(input_attr_name)
self.assertEqual(len(pipeline._pipelined_postprocs[1]._args.args), 1)
self.assertEqual(len(pipeline._pipelined_postprocs[1]._args.kwargs), 0)
self.assertEqual(len(pipeline._pipelined_postprocs[1]._args.args[0].steps), 2)
self.assertEqual(
pipeline._pipelined_postprocs[1]._args.args[0].steps[0], NoopArgInfoStep()
)
self.assertIsInstance(
pipeline._pipelined_postprocs[1]._args.args[0].steps[1], GetAttrArgInfoStep
)

self.assertEqual(postproc_mod._args[0].is_getitems, [False, False])
# no parent postproc module in FX graph
self.assertEqual(postproc_mod._args[0].postproc_modules, [None, None])
get_arg_infos = {
# pyre-fixme[16]: assertions above ensure that steps[1] is a GetAttrArgInfoStep
postproc._args.args[0].steps[1].attr_name
for postproc in pipeline._pipelined_postprocs
}
self.assertEqual(get_arg_infos, {"idlist_features", "idscore_features"})

# pyre-ignore
@unittest.skipIf(
Expand Down Expand Up @@ -1235,69 +1251,63 @@ def test_pipeline_postproc_recursive(self) -> None:
pipelined_weighted_ebc = pipeline._pipelined_modules[1]

# Check pipelined args
for ebc in [pipelined_ebc, pipelined_weighted_ebc]:
self.assertEqual(len(ebc.forward._args), 1)
self.assertEqual(ebc.forward._args[0].input_attrs, ["", 0])
self.assertEqual(ebc.forward._args[0].is_getitems, [False, True])
self.assertEqual(len(ebc.forward._args[0].postproc_modules), 2)
self.assertIsInstance(
ebc.forward._args[0].postproc_modules[0], PipelinedPostproc
)
self.assertEqual(ebc.forward._args[0].postproc_modules[1], None)

self.assertEqual(len(pipelined_ebc.forward._args.args), 1)
self.assertEqual(len(pipelined_ebc.forward._args.kwargs), 0)
self.assertEqual(
pipelined_ebc.forward._args[0].postproc_modules[0],
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_nonweighted`.
pipelined_model.module.postproc_nonweighted,
pipelined_ebc.forward._args.args[0].steps,
[
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `_postproc_module`.
PostprocArgInfoStep(pipelined_model.module.postproc_nonweighted),
GetItemArgInfoStep(0),
],
)
self.assertEqual(len(pipelined_weighted_ebc.forward._args.args), 1)
self.assertEqual(len(pipelined_weighted_ebc.forward._args.kwargs), 0)
self.assertEqual(
pipelined_weighted_ebc.forward._args[0].postproc_modules[0],
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_weighted`.
pipelined_model.module.postproc_weighted,
pipelined_weighted_ebc.forward._args.args[0].steps,
[
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `_postproc_module`.
PostprocArgInfoStep(pipelined_model.module.postproc_weighted),
GetItemArgInfoStep(0),
],
)

# postproc args
self.assertEqual(len(pipeline._pipelined_postprocs), 3)

# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `_postproc_module`.
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `_postproc_module`.
parent_postproc_mod = pipelined_model.module._postproc_module

for postproc_mod in pipeline._pipelined_postprocs:
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_nonweighted`.
if postproc_mod == pipelined_model.module.postproc_nonweighted:
self.assertEqual(len(postproc_mod._args), 1)
args = postproc_mod._args[0]
self.assertEqual(args.input_attrs, ["", "idlist_features"])
self.assertEqual(args.is_getitems, [False, False])
self.assertEqual(len(args.postproc_modules), 2)
self.assertEqual(len(postproc_mod._args.args), 1)
self.assertEqual(len(postproc_mod._args.kwargs), 0)
self.assertEqual(
args.postproc_modules[0],
parent_postproc_mod,
postproc_mod._args.args[0].steps,
[
PostprocArgInfoStep(parent_postproc_mod),
GetAttrArgInfoStep("idlist_features"),
],
)
self.assertEqual(args.postproc_modules[1], None)

# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_weighted`.
elif postproc_mod == pipelined_model.module.postproc_weighted:
self.assertEqual(len(postproc_mod._args), 1)
args = postproc_mod._args[0]
self.assertEqual(args.input_attrs, ["", "idscore_features"])
self.assertEqual(args.is_getitems, [False, False])
self.assertEqual(len(args.postproc_modules), 2)
self.assertEqual(len(postproc_mod._args.args), 1)
self.assertEqual(len(postproc_mod._args.kwargs), 0)
self.assertEqual(
args.postproc_modules[0],
parent_postproc_mod,
postproc_mod._args.args[0].steps,
[
PostprocArgInfoStep(parent_postproc_mod),
GetAttrArgInfoStep("idscore_features"),
],
)
self.assertEqual(args.postproc_modules[1], None)
elif postproc_mod == parent_postproc_mod:
self.assertEqual(len(postproc_mod._args), 1)
args = postproc_mod._args[0]
self.assertEqual(args.input_attrs, [""])
self.assertEqual(args.is_getitems, [False])
self.assertEqual(args.postproc_modules, [None])
self.assertEqual(len(postproc_mod._args.args), 1)
self.assertEqual(len(postproc_mod._args.kwargs), 0)
self.assertEqual(postproc_mod._args.args[0].steps, [NoopArgInfoStep()])

# pyre-ignore
@unittest.skipIf(
Expand Down
112 changes: 42 additions & 70 deletions torchrec/distributed/train_pipeline/tests/test_train_pipelines_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,10 +23,11 @@
TrainPipelineSparseDistTestBase,
)
from torchrec.distributed.train_pipeline.utils import (
_build_args_kwargs,
_get_node_args,
_rewrite_model,
ArgInfo,
ArgInfoStepFactory,
CallArgs,
NodeArgsHelper,
PipelinedForward,
PipelinedPostproc,
TrainPipelineContext,
Expand Down Expand Up @@ -110,17 +111,19 @@ def test_rewrite_model(self) -> None:
self.assertEqual(
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `sparse`.
sharded_model.module.sparse.ebc.forward._args[0].postproc_modules[0],
sharded_model.module.sparse.ebc.forward._args.args[0]
.steps[0]
.postproc_module,
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_module`.
sharded_model.module.postproc_module,
)
self.assertEqual(
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `sparse`.
sharded_model.module.sparse.weighted_ebc.forward._args[0].postproc_modules[
0
],
sharded_model.module.sparse.weighted_ebc.forward._args.args[0]
.steps[0]
.postproc_module,
# pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute
# `postproc_module`.
sharded_model.module.postproc_module,
Expand Down Expand Up @@ -154,7 +157,7 @@ def forward(self, x):
rewritten_model.test_module = PipelinedPostproc(
postproc_module=rewritten_model.test_module,
fqn="test_module",
args=[],
args=CallArgs(args=[], kwargs={}),
context=TrainPipelineContext(),
default_stream=MagicMock(),
dist_stream=MagicMock(),
Expand Down Expand Up @@ -260,83 +263,53 @@ def test_restore_from_snapshot(self) -> None:
@parameterized.expand(
[
(
[
# Empty attrs to ignore any attr based logic.
ArgInfo(
input_attrs=[
"",
],
is_getitems=[False],
postproc_modules=[None],
constants=[None],
name="id_list_features",
),
ArgInfo(
input_attrs=[],
is_getitems=[],
postproc_modules=[],
constants=[],
name="id_score_list_features",
),
],
CallArgs(
args=[],
kwargs={
"id_list_features": ArgInfo(steps=[ArgInfoStepFactory.noop()]),
# Empty attrs to ignore any attr based logic.
"id_score_list_features": ArgInfo(
steps=[ArgInfoStepFactory.noop()]
),
},
),
0,
["id_list_features", "id_score_list_features"],
),
(
[
# Empty attrs to ignore any attr based logic.
ArgInfo(
input_attrs=[
"",
],
is_getitems=[False],
postproc_modules=[None],
constants=[None],
name=None,
),
ArgInfo(
input_attrs=[],
is_getitems=[],
postproc_modules=[],
constants=[],
name=None,
),
],
CallArgs(
args=[
# Empty attrs to ignore any attr based logic.
ArgInfo(steps=[ArgInfoStepFactory.noop()]),
ArgInfo(steps=[]),
],
kwargs={},
),
2,
[],
),
(
[
# Empty attrs to ignore any attr based logic.
ArgInfo(
input_attrs=[
"",
],
is_getitems=[False],
postproc_modules=[None],
constants=[None],
name=None,
),
ArgInfo(
input_attrs=[],
is_getitems=[],
postproc_modules=[],
constants=[],
name="id_score_list_features",
),
],
CallArgs(
args=[
# Empty attrs to ignore any attr based logic.
ArgInfo(
steps=[ArgInfoStepFactory.noop()],
)
],
kwargs={"id_score_list_features": ArgInfo(steps=[])},
),
1,
["id_score_list_features"],
),
]
)
def test_build_args_kwargs(
self,
fwd_args: List[ArgInfo],
fwd_args: CallArgs,
args_len: int,
kwarges_keys: List[str],
) -> None:
args, kwargs = _build_args_kwargs("initial_input", fwd_args)
args, kwargs = fwd_args.build_args_kwargs("initial_input")
self.assertEqual(len(args), args_len)
self.assertEqual(list(kwargs.keys()), kwarges_keys)

Expand Down Expand Up @@ -367,10 +340,9 @@ def test_get_node_args_helper_call_module_kjt(self) -> None:
{},
)

num_found = 0
_, num_found = _get_node_args(
MagicMock(), kjt_node, set(), TrainPipelineContext(), False
)
node_args_helper = NodeArgsHelper(MagicMock(), TrainPipelineContext(), False)

_, num_found = node_args_helper.get_node_args(kjt_node)

# Weights is call_module node, so we should only find 2 args unmodified
self.assertEqual(num_found, len(kjt_args) - 1)
Loading
Loading