Skip to content

Use tf.shape for graph mode support #6

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

Merged
merged 2 commits into from
Feb 23, 2023
Merged

Use tf.shape for graph mode support #6

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a50037f
Use tf.shape for graph mode support
edknv Feb 14, 2023
bf76a45
Fall back to tf.shape only when static shape is not available
edknv Feb 22, 2023
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
10 changes: 7 additions & 3 deletions distributed_embeddings/python/layers/dist_model_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -395,17 +395,21 @@ def _call_base(self, inputs): # pylint: disable=missing-param-doc,missing-type-
local_shapes, local_splits, global_splits, flat_inputs = [], [], [], []
for rank_input_ids in self.strategy.input_ids_list:
rank_inputs = [inputs[index] for index in rank_input_ids]
local_shapes.append([inp.shape for inp in rank_inputs])
local_shapes.append(
[tf.shape(inp) if None in inp.shape else inp.shape for inp in rank_inputs]
)
rank_inputs = [tf.reshape(inp, [-1]) for inp in rank_inputs]
local_splits.append([inp.shape[0] for inp in rank_inputs])
local_splits.append(
[tf.shape(inp)[0] if inp.shape[0] is None else inp.shape[0] for inp in rank_inputs]
)
global_splits.append(sum(local_splits[-1]))
flat_inputs += rank_inputs
inputs = tf.concat(flat_inputs, 0)
inputs, _ = hvd.alltoall(inputs, splits=global_splits, name='inp_dp_to_mp')
inputs = tf.reshape(inputs, [self.world_size, -1])
inputs = tf.split(inputs, local_splits[self.rank], 1)
inputs = [
tf.reshape(inp, [self.world_size * shape[0]] + shape[1:])
tf.reshape(inp, tf.concat([[self.world_size * shape[0]], shape[1:]], 0))
for inp, shape in zip(inputs, local_shapes[self.rank])
]
else:
Expand Down
2 changes: 1 addition & 1 deletion tests/dist_model_parallel_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -274,7 +274,7 @@ def test_column_slice_merge(self):

dp_inputs, _ = self.gen_inputs(table_sizes)
self.run_and_test(ref_model, dp_inputs, test_model, dp_inputs)
for tables in test_model.dist_embeddings.strategy.table_ids_list:
for tables in test_model.dist_embeddings.strategy.table_ids:
self.assertEqual(len(tables), len(set(tables)))

def test_column_slice_threshold(self):
Expand Down