You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by mb...@apache.org on 2021/03/16 17:39:32 UTC
[tvm] branch revert-7666-onnx_nms_empty_init created (now 10d4341)
This is an automated email from the ASF dual-hosted git repository.
mbrookhart pushed a change to branch revert-7666-onnx_nms_empty_init
in repository https://gitbox.apache.org/repos/asf/tvm.git.
at 10d4341 Revert "init the concat tensor with 1s and then slice them away (#7666)"
This branch includes the following new commits:
new 10d4341 Revert "init the concat tensor with 1s and then slice them away (#7666)"
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
[tvm] 01/01: Revert "init the concat tensor with 1s and then slice
them away (#7666)"
Posted by mb...@apache.org.
This is an automated email from the ASF dual-hosted git repository.
mbrookhart pushed a commit to branch revert-7666-onnx_nms_empty_init
in repository https://gitbox.apache.org/repos/asf/tvm.git
commit 10d4341d10dbd86ec7ab5c10a734427e07d92653
Author: Matthew Brookhart <mb...@octoml.ai>
AuthorDate: Tue Mar 16 11:39:11 2021 -0600
Revert "init the concat tensor with 1s and then slice them away (#7666)"
This reverts commit 348d4e7532f7d5ce68d76e7634a5f11b7dc60ab8.
---
python/tvm/relay/frontend/onnx.py | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/python/tvm/relay/frontend/onnx.py b/python/tvm/relay/frontend/onnx.py
index 391eaaa..4a0a1ed 100644
--- a/python/tvm/relay/frontend/onnx.py
+++ b/python/tvm/relay/frontend/onnx.py
@@ -2629,10 +2629,10 @@ class NonMaxSuppression(OnnxOpConverter):
# Call the second loop, rework outputs into correct form
init_count = _op.const(np.array([0]).astype("int64"), dtype="int64")
- init_out = _op.const(np.array([1, 1, 1]).reshape([1, 3]).astype("int64"), dtype="int64")
+ init_out = _op.const(np.array([]).reshape([0, 3]).astype("int64"), dtype="int64")
loop_vals = outer_loop(init_count, B, C, onnx_output, nms_size_output, init_out)
- loop_out = _expr.TupleGetItem(loop_vals, 5)
- return _op.strided_slice(loop_out, [1, 0], shape_of(loop_out), [1, 1])
+
+ return _expr.TupleGetItem(loop_vals, 5)
# compatible operators that do NOT require any conversion.