Skip to content

A fix for seed attribute in the keras random normal generator #2126

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 8 commits into from
Mar 16, 2023
Merged

A fix for seed attribute in the keras random normal generator #2126

Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
599a424
A fix for seed field in the tf.keras random normal generator. The see…
inonbe Feb 23, 2023
0fca209
Merge branch 'main' into random_seed
fatcat-z Mar 14, 2023
f46f999
update version_1 for const shape as well.
inonbe Mar 14, 2023
cc2193b
update version_1 method for constant shapes as well.
inonbe Mar 14, 2023
a6e6b2b
update version_1 method for constant shapes as well.
inonbe Mar 14, 2023
bdae6f1
Merge remote-tracking branch 'origin/random_seed' into random_seed
inonbe Mar 14, 2023
a206179
Merge branch 'main' into random_seed
fatcat-z Mar 15, 2023
0960fcf
Merge branch 'main' into random_seed
fatcat-z Mar 16, 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
8 changes: 4 additions & 4 deletions tf2onnx/onnx_opset/generator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,8 +60,8 @@ def version_1(cls, ctx, node, **kwargs):
# in the rewriter does not trigger. grappler will send the random uniform
# with shape as input so we need to pickup the input here and if the shape is
# const we make it an attribute.
seed = node.get_attr("seed")
node.set_attr("seed", float(seed.f))
seed = node.get_attr("seed2")
node.set_attr("seed", float(seed.i))
utils.make_sure(node.inputs[0].is_const(), "%s node with non-const shape requires opset >= 9", node.type)
shape = node.inputs[0].get_tensor_value()
ctx.remove_input(node, node.input[0], 0)
Expand All @@ -88,8 +88,8 @@ def version_9(cls, ctx, node, **kwargs):
if node.inputs[0].is_const():
cls.version_1(ctx, node, **kwargs)
else:
seed = node.get_attr("seed")
node.set_attr("seed", float(seed.f))
seed = node.get_attr("seed2")
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need to also update this for version_1(line 63)?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks @fatcat-z you're quite right, ill update version_1 as well.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm a little confused: is this behavior (using seed2 instead of seed) same among different TF versions? Did it change after one of tf version?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As far as I've seen and tested, version_1 behaviour is also relevant for TF version.2
when the random generator dim/shape are constant and not derived from the data
batch size or any other dependent size.

I'm checking with TF V1 whether it was changed.

node.set_attr("seed", float(seed.i))
cast_node = ctx.make_node("Cast", [node.input[0]], attr={'to': onnx_pb.TensorProto.INT64})
const_node = ctx.make_node("ConstantOfShape", cast_node.output)
inputs = node.input.copy()
Expand Down