This is an automated email from the ASF dual-hosted git repository.
andrewzhaoluo pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new 5400b942d2 [ONNX] Add converter for FastGelu from Microsoft
onnxruntime contrib opset (#13119)
5400b942d2 is described below
commit 5400b942d2ef5ec7145f4f3e253b24c8abb4265e
Author: Alexey Gladyshev <[email protected]>
AuthorDate: Mon Oct 24 20:00:27 2022 +0300
[ONNX] Add converter for FastGelu from Microsoft onnxruntime contrib opset
(#13119)
* add converter for FastGelu from Microsoft onnxruntime contrib opset
* integrate FastGelu into test system for ONNX converters
* code review fixes
* returned constant calculation
---
python/tvm/relay/frontend/onnx.py | 38 +++++++++++++++++-
tests/python/frontend/onnx/test_forward.py | 62 ++++++++++++++++++++----------
2 files changed, 78 insertions(+), 22 deletions(-)
diff --git a/python/tvm/relay/frontend/onnx.py
b/python/tvm/relay/frontend/onnx.py
index 743e39296c..ca290cf9a8 100644
--- a/python/tvm/relay/frontend/onnx.py
+++ b/python/tvm/relay/frontend/onnx.py
@@ -928,10 +928,45 @@ class Gelu(OnnxOpConverter):
return _op.multiply(term1, term2)
+class FastGelu(OnnxOpConverter):
+ """Operator converter for FastGelu from Microsoft onnxruntime contrib
opset.
+
+ fast_gelu(x) = 0.5x(1 + tanh(sqrt(2/pi)(x + 0.044715x^3)))
+ = 0.5x(1 + tanh((sqrt(2/pi)x + 0.044715(sqrt(2/pi)x^3)))
+ = 0.5x(1 + tanh(c1 * x + c2 * x^3)))
+ , where
+ c1 = sqrt(2/pi)
+ c2 = 0.044715 * sqrt(2/pi)
+ """
+
+ @classmethod
+ def _impl_v1(cls, inputs, attr, params):
+ x = inputs[0]
+ if inputs[1]:
+ bias = inputs[1]
+ bias_shape = infer_shape(bias)
+ assert len(bias_shape) == 1, "bias term must be a 1D tensor"
+ x += bias
+
+ # Declare consts
+ const_dtype = infer_type(x).checked_type.dtype
+ half = _expr.const(0.5, dtype=const_dtype)
+ one = _expr.const(1.0, dtype=const_dtype)
+ const1 = _expr.const(math.sqrt(2 / math.pi), dtype=const_dtype)
+ const2 = _expr.const(0.044715 * math.sqrt(2 / math.pi),
dtype=const_dtype)
+
+ # Compute FastGelu
+ term1 = _op.multiply(half, x)
+ term2 = _op.multiply(const1, x)
+ term3 = _op.multiply(const2, _op.power(x, _expr.const(3, const_dtype)))
+ tanh = _op.tanh(_op.add(term2, term3))
+ return _op.multiply(term1, _op.add(one, tanh))
+
+
class BiasGelu(OnnxOpConverter):
"""Operator converter for BiasGelu from Microsoft onnxruntime contrib
opset.
- bias_gelu(x, b) = 0.5(x, b)(1 + erf((x + b)/sqrt(2)))
+ bias_gelu(x, b) = 0.5(x + b)(1 + erf((x + b)/sqrt(2)))
"""
@classmethod
@@ -5384,6 +5419,7 @@ def _get_convert_map(opset):
"Selu": Selu.get_converter(opset),
"Elu": Elu.get_converter(opset),
"Gelu": Gelu.get_converter(opset),
+ "FastGelu": FastGelu.get_converter(opset),
"BiasGelu": BiasGelu.get_converter(opset),
"LayerNormalization": LayerNormalization.get_converter(opset),
# TODO: We need a better way to handle different domains, in case
diff --git a/tests/python/frontend/onnx/test_forward.py
b/tests/python/frontend/onnx/test_forward.py
index 684a33fbce..5eac9a8dd4 100644
--- a/tests/python/frontend/onnx/test_forward.py
+++ b/tests/python/frontend/onnx/test_forward.py
@@ -5614,13 +5614,18 @@ def test_reverse_sequence(target, dev):
verify_reverse_sequence(x, sequence_lens, 1, 0)
[email protected]("op_name", ["Gelu", "FastGelu"], scope="session")
[email protected]("data_type", ["float16", "float32"], scope="session")
@tvm.testing.parametrize_targets
-def test_gelu(target, dev):
+def test_gelu(target, dev, data_type, op_name):
"""test_gelu"""
+ dtype = np.dtype(data_type)
+ tensor_type = mapping.NP_TYPE_TO_TENSOR_TYPE[dtype]
+ absolute_tolerance = 1e-3 if data_type == "float16" else 1e-5
def verify_gelu(x):
node = onnx.helper.make_node(
- "Gelu",
+ op_name,
inputs=["x"],
outputs=["y"],
domain="com.microsoft",
@@ -5628,27 +5633,34 @@ def test_gelu(target, dev):
graph = helper.make_graph(
[node],
- "gelu_test",
- inputs=[helper.make_tensor_value_info("x", TensorProto.FLOAT,
list(x.shape))],
- outputs=[helper.make_tensor_value_info("y", TensorProto.FLOAT,
list(x.shape))],
+ f"{op_name}_test",
+ inputs=[helper.make_tensor_value_info("x", tensor_type,
list(x.shape))],
+ outputs=[helper.make_tensor_value_info("y", tensor_type,
list(x.shape))],
)
- model = helper.make_model(graph, producer_name="gelu_test")
- verify_with_ort_with_inputs(model, [x], [x.shape], target=target,
dev=dev)
+ model = helper.make_model(graph, producer_name=f"{op_name}_test")
+ verify_with_ort_with_inputs(
+ model, [x], [x.shape], atol=absolute_tolerance, dtype=data_type,
target=target, dev=dev
+ )
- x = np.array([-1.0, 0, 1.0, 100.0, -100.0, 1000.0, -1000.0],
dtype=np.float32)
+ x = np.array([-1.0, 0, 1.0, 100.0, -100.0, 1000.0, -1000.0], dtype=dtype)
verify_gelu(x)
- x = np.array([[1, 2], [3, 4]], dtype=np.float32)
+ x = np.array([[1, 2], [3, 4]], dtype=dtype)
verify_gelu(x)
[email protected]("op_name", ["BiasGelu", "FastGelu"], scope="session")
[email protected]("data_type", ["float16", "float32"], scope="session")
@tvm.testing.parametrize_targets
-def test_biasgelu(target, dev):
+def test_biasgelu(target, dev, data_type, op_name):
"""test_biasgelu"""
+ dtype = np.dtype(data_type)
+ tensor_type = mapping.NP_TYPE_TO_TENSOR_TYPE[dtype]
+ absolute_tolerance = 1e-3 if data_type == "float16" else 1e-5
def verify_biasgelu(x, bias):
node = onnx.helper.make_node(
- "BiasGelu",
+ op_name,
inputs=["x", "bias"],
outputs=["y"],
domain="com.microsoft",
@@ -5656,23 +5668,31 @@ def test_biasgelu(target, dev):
graph = helper.make_graph(
[node],
- "biasgelu_test",
+ f"{op_name}_test",
inputs=[
- helper.make_tensor_value_info("x", TensorProto.FLOAT,
list(x.shape)),
- helper.make_tensor_value_info("bias", TensorProto.FLOAT,
list(bias.shape)),
+ helper.make_tensor_value_info("x", tensor_type, list(x.shape)),
+ helper.make_tensor_value_info("bias", tensor_type,
list(bias.shape)),
],
- outputs=[helper.make_tensor_value_info("y", TensorProto.FLOAT,
list(x.shape))],
+ outputs=[helper.make_tensor_value_info("y", tensor_type,
list(x.shape))],
)
- model = helper.make_model(graph, producer_name="biasgelu_test")
- verify_with_ort_with_inputs(model, [x, bias], [x.shape],
target=target, dev=dev)
+ model = helper.make_model(graph, producer_name=f"{op_name}_test")
+ verify_with_ort_with_inputs(
+ model,
+ [x, bias],
+ [x.shape],
+ atol=absolute_tolerance,
+ dtype=data_type,
+ target=target,
+ dev=dev,
+ )
- x = np.array([-1.0, 0, 1.0, 100.0, -100.0, 1000.0, -1000.0],
dtype=np.float32)
- bias = np.repeat(2.0, 7).astype("float32")
+ x = np.array([-1.0, 0, 1.0, 100.0, -100.0, 1000.0, -1000.0], dtype=dtype)
+ bias = np.repeat(2.0, 7).astype(dtype)
verify_biasgelu(x, bias)
- x = np.array([[1, 2], [3, 4]], dtype=np.float32)
- bias = np.array([0.3, 4.0], dtype=np.float32)
+ x = np.array([[1, 2], [3, 4]], dtype=dtype)
+ bias = np.array([0.3, 4.0], dtype=dtype)
verify_biasgelu(x, bias)
