Skip to content

Export the cosine_similarity op as an ATenOp correctly #21884

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

Closed
houseroad wants to merge 1 commit into from
Closed

Export the cosine_similarity op as an ATenOp correctly #21884

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
7 changes: 7 additions & 0 deletions test/onnx/test_pytorch_onnx_caffe2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -891,6 +891,13 @@ def forward(self, x):
x = torch.randn(*shape)
self.run_model_test(MyModel(), train=False, input=(x,), batch_size=BATCH_SIZE, use_gpu=False)

def test_cosine_similarity(self):
shape = (100, 128)
x = torch.randn(*shape)
y = torch.randn(*shape)
self.run_model_test(torch.nn.CosineSimilarity(dim=1, eps=1e-6), train=False,
input=(x, y), batch_size=BATCH_SIZE, use_gpu=False)

def test_lstm_constant_folding(self):
class LstmNet(nn.Module):
def __init__(self, input_size, hidden_size, num_layers, bidirectional):
Expand Down
5 changes: 5 additions & 0 deletions torch/onnx/symbolic_opset9.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -958,6 +958,11 @@ def layer_norm(g, self, normalized_shape, weight, bias, eps, cudnn_enable):
eps_f=eps, cudnn_enable_i=cudnn_enable, operator_s="layer_norm")


@parse_args('v', 'v', 'i', 'f')
def cosine_similarity(g, x1, x2, dim, eps):
return g.op("ATen", x1, x2, dim_i=dim, eps_f=eps, operator_s="cosine_similarity")


# ignore clone operators that are inserted by PyTorch autograd
def clone(g, input):
return input
Expand Down