You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/02/18 18:28:15 UTC
[GitHub] [tvm] ANSHUMAN87 opened a new pull request #7473: [Frontend] TF V2 sparse.todense() test added
ANSHUMAN87 opened a new pull request #7473:
URL: https://github.com/apache/tvm/pull/7473
tf.sparse_to_dense() is deprecated. TF V2 sparse.todense() test added.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] mbaret merged pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
mbaret merged pull request #7473:
URL: https://github.com/apache/tvm/pull/7473
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ANSHUMAN87 commented on pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
ANSHUMAN87 commented on pull request #7473:
URL: https://github.com/apache/tvm/pull/7473#issuecomment-781829886
cc @tkonolige , @FrozenGene , @zhiics !
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ANSHUMAN87 commented on a change in pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
ANSHUMAN87 commented on a change in pull request #7473:
URL: https://github.com/apache/tvm/pull/7473#discussion_r582148333
##########
File path: tests/python/frontend/tensorflow/test_forward.py
##########
@@ -1915,6 +1915,105 @@ def test_forward_sparse_fill_empty_rows(
)
+#######################################################################
+# tensorflow.compat.v1.sparse_to_dense
+# ---------------
+def _test_sparse_to_dense(sparse_indices, sparse_values, default_value, output_shape):
+ with tf.Graph().as_default():
+ indices = tf.placeholder(
+ shape=sparse_indices.shape, dtype=str(sparse_indices.dtype), name="indices"
+ )
+ values = tf.placeholder(
+ shape=sparse_values.shape, dtype=str(sparse_values.dtype), name="values"
+ )
+ oshape = tf.constant(output_shape, shape=output_shape.shape, dtype=str(output_shape.dtype))
+
+ if default_value == None:
+ output = tf.sparse_to_dense(indices, oshape, values)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values], ["indices:0", "values:0"], output.name
+ )
+ else:
+ dv = tf.placeholder(shape=(), dtype=str(default_value.dtype), name="default_value")
+ output = tf.sparse_to_dense(indices, oshape, values, dv)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values, default_value],
+ ["indices:0", "values:0", "default_value:0"],
+ output.name,
+ )
+
+
+def test_forward_sparse_to_dense():
+ # scalar
+ _test_sparse_to_dense(
+ sparse_indices=np.int32(1),
+ sparse_values=np.int32(3),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3, 3, 3]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector nXd
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0], [1, 2]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([3, 4]).astype("int32"),
+ )
+
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0, 0], [1, 2, 3]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(4),
+ output_shape=np.array([2, 3, 4]).astype("int32"),
+ )
+
+ # floats
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=np.float32(3.5),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # default value not specified
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=None,
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+
+#######################################################################
+# tensorflow.sparse.to_dense
+# ---------------
+def _test_sparse_to_dense_v2(indices, values, A_shape, dtype, default_value=None):
+ with tf.Graph().as_default():
+ A_sp = tf.sparse.SparseTensor(indices=indices, values=values, dense_shape=A_shape)
+ B = tf.placeholder(shape=A_shape, dtype=dtype, name="B")
+
+ result = tf.sparse.to_dense(A_sp, default_value=default_value) + B
+
+ B_np = np.random.uniform(high=5.0, size=A_shape).astype(dtype)
+
+ compare_tf_with_tvm([B_np], [B.name], result.name)
+
+
+def test_forward_sparse_to_dense_v2():
+ _test_sparse_to_dense_v2([[0, 0], [1, 2]], [4.0, 8.0], [3, 4], "float32")
Review comment:
Thanks for pointing. scalar testcase is added now.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ANSHUMAN87 commented on a change in pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
ANSHUMAN87 commented on a change in pull request #7473:
URL: https://github.com/apache/tvm/pull/7473#discussion_r582148802
##########
File path: tests/python/frontend/tensorflow/test_forward.py
##########
@@ -1915,6 +1915,105 @@ def test_forward_sparse_fill_empty_rows(
)
+#######################################################################
+# tensorflow.compat.v1.sparse_to_dense
+# ---------------
+def _test_sparse_to_dense(sparse_indices, sparse_values, default_value, output_shape):
+ with tf.Graph().as_default():
+ indices = tf.placeholder(
+ shape=sparse_indices.shape, dtype=str(sparse_indices.dtype), name="indices"
+ )
+ values = tf.placeholder(
+ shape=sparse_values.shape, dtype=str(sparse_values.dtype), name="values"
+ )
+ oshape = tf.constant(output_shape, shape=output_shape.shape, dtype=str(output_shape.dtype))
+
+ if default_value == None:
+ output = tf.sparse_to_dense(indices, oshape, values)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values], ["indices:0", "values:0"], output.name
+ )
+ else:
+ dv = tf.placeholder(shape=(), dtype=str(default_value.dtype), name="default_value")
+ output = tf.sparse_to_dense(indices, oshape, values, dv)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values, default_value],
+ ["indices:0", "values:0", "default_value:0"],
+ output.name,
+ )
+
+
+def test_forward_sparse_to_dense():
+ # scalar
+ _test_sparse_to_dense(
+ sparse_indices=np.int32(1),
+ sparse_values=np.int32(3),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3, 3, 3]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector nXd
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0], [1, 2]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([3, 4]).astype("int32"),
+ )
+
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0, 0], [1, 2, 3]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(4),
+ output_shape=np.array([2, 3, 4]).astype("int32"),
+ )
+
+ # floats
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=np.float32(3.5),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # default value not specified
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=None,
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+
+#######################################################################
+# tensorflow.sparse.to_dense
+# ---------------
+def _test_sparse_to_dense_v2(indices, values, A_shape, dtype, default_value=None):
+ with tf.Graph().as_default():
+ A_sp = tf.sparse.SparseTensor(indices=indices, values=values, dense_shape=A_shape)
+ B = tf.placeholder(shape=A_shape, dtype=dtype, name="B")
+
+ result = tf.sparse.to_dense(A_sp, default_value=default_value) + B
Review comment:
Updated now. PTAL. Thanks!
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] mbaret commented on pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
mbaret commented on pull request #7473:
URL: https://github.com/apache/tvm/pull/7473#issuecomment-785768364
Thanks @ANSHUMAN87, this is now merged.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] mbaret commented on a change in pull request #7473: [Frontend] TF V2 sparse.todense() test added
Posted by GitBox <gi...@apache.org>.
mbaret commented on a change in pull request #7473:
URL: https://github.com/apache/tvm/pull/7473#discussion_r580658725
##########
File path: tests/python/frontend/tensorflow/test_forward.py
##########
@@ -1915,6 +1915,105 @@ def test_forward_sparse_fill_empty_rows(
)
+#######################################################################
+# tensorflow.compat.v1.sparse_to_dense
+# ---------------
+def _test_sparse_to_dense(sparse_indices, sparse_values, default_value, output_shape):
+ with tf.Graph().as_default():
+ indices = tf.placeholder(
+ shape=sparse_indices.shape, dtype=str(sparse_indices.dtype), name="indices"
+ )
+ values = tf.placeholder(
+ shape=sparse_values.shape, dtype=str(sparse_values.dtype), name="values"
+ )
+ oshape = tf.constant(output_shape, shape=output_shape.shape, dtype=str(output_shape.dtype))
+
+ if default_value == None:
+ output = tf.sparse_to_dense(indices, oshape, values)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values], ["indices:0", "values:0"], output.name
+ )
+ else:
+ dv = tf.placeholder(shape=(), dtype=str(default_value.dtype), name="default_value")
+ output = tf.sparse_to_dense(indices, oshape, values, dv)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values, default_value],
+ ["indices:0", "values:0", "default_value:0"],
+ output.name,
+ )
+
+
+def test_forward_sparse_to_dense():
+ # scalar
+ _test_sparse_to_dense(
+ sparse_indices=np.int32(1),
+ sparse_values=np.int32(3),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3, 3, 3]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector nXd
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0], [1, 2]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([3, 4]).astype("int32"),
+ )
+
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0, 0], [1, 2, 3]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(4),
+ output_shape=np.array([2, 3, 4]).astype("int32"),
+ )
+
+ # floats
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=np.float32(3.5),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # default value not specified
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=None,
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+
+#######################################################################
+# tensorflow.sparse.to_dense
+# ---------------
+def _test_sparse_to_dense_v2(indices, values, A_shape, dtype, default_value=None):
+ with tf.Graph().as_default():
+ A_sp = tf.sparse.SparseTensor(indices=indices, values=values, dense_shape=A_shape)
+ B = tf.placeholder(shape=A_shape, dtype=dtype, name="B")
+
+ result = tf.sparse.to_dense(A_sp, default_value=default_value) + B
Review comment:
What's the use of the '+ B' here (does the compiler otherwise optimize the operator away)?
##########
File path: tests/python/frontend/tensorflow/test_forward.py
##########
@@ -1915,6 +1915,105 @@ def test_forward_sparse_fill_empty_rows(
)
+#######################################################################
+# tensorflow.compat.v1.sparse_to_dense
+# ---------------
+def _test_sparse_to_dense(sparse_indices, sparse_values, default_value, output_shape):
+ with tf.Graph().as_default():
+ indices = tf.placeholder(
+ shape=sparse_indices.shape, dtype=str(sparse_indices.dtype), name="indices"
+ )
+ values = tf.placeholder(
+ shape=sparse_values.shape, dtype=str(sparse_values.dtype), name="values"
+ )
+ oshape = tf.constant(output_shape, shape=output_shape.shape, dtype=str(output_shape.dtype))
+
+ if default_value == None:
+ output = tf.sparse_to_dense(indices, oshape, values)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values], ["indices:0", "values:0"], output.name
+ )
+ else:
+ dv = tf.placeholder(shape=(), dtype=str(default_value.dtype), name="default_value")
+ output = tf.sparse_to_dense(indices, oshape, values, dv)
+ compare_tf_with_tvm(
+ [sparse_indices, sparse_values, default_value],
+ ["indices:0", "values:0", "default_value:0"],
+ output.name,
+ )
+
+
+def test_forward_sparse_to_dense():
+ # scalar
+ _test_sparse_to_dense(
+ sparse_indices=np.int32(1),
+ sparse_values=np.int32(3),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3, 3, 3]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # vector nXd
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0], [1, 2]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(0),
+ output_shape=np.array([3, 4]).astype("int32"),
+ )
+
+ _test_sparse_to_dense(
+ sparse_indices=np.array([[0, 0, 0], [1, 2, 3]]).astype("int32"),
+ sparse_values=np.array([1, 2]).astype("int32"),
+ default_value=np.int32(4),
+ output_shape=np.array([2, 3, 4]).astype("int32"),
+ )
+
+ # floats
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=np.float32(3.5),
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+ # default value not specified
+ _test_sparse_to_dense(
+ sparse_indices=np.array([0, 1, 4]).astype("int32"),
+ sparse_values=np.array([3.1, 3.1, 3.1]).astype("float32"),
+ default_value=None,
+ output_shape=np.array([5]).astype("int32"),
+ )
+
+
+#######################################################################
+# tensorflow.sparse.to_dense
+# ---------------
+def _test_sparse_to_dense_v2(indices, values, A_shape, dtype, default_value=None):
+ with tf.Graph().as_default():
+ A_sp = tf.sparse.SparseTensor(indices=indices, values=values, dense_shape=A_shape)
+ B = tf.placeholder(shape=A_shape, dtype=dtype, name="B")
+
+ result = tf.sparse.to_dense(A_sp, default_value=default_value) + B
+
+ B_np = np.random.uniform(high=5.0, size=A_shape).astype(dtype)
+
+ compare_tf_with_tvm([B_np], [B.name], result.name)
+
+
+def test_forward_sparse_to_dense_v2():
+ _test_sparse_to_dense_v2([[0, 0], [1, 2]], [4.0, 8.0], [3, 4], "float32")
Review comment:
Is there a reason we don't replicate the test cases of the 'old' operator (e.g. scalar test)?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org