Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Supports float 8 initializers in ReferenceEvaluator #5295

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 5 commits into from
Jun 15, 2023
+27 −3
Merged

Supports float 8 initializers in ReferenceEvaluator #5295

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
77ac8ea
Supports float 8 initializers in ReferenceEvaluator
xadupre Jun 9, 2023
f4ee930
remove unnecessary prints
xadupre Jun 9, 2023
72f1803
remove unused import
xadupre Jun 9, 2023
b9fb1a0
Merge branch 'main' into ref8
jcwchen Jun 13, 2023
215f2ea
Merge branch 'main' of https://github.com/onnx/onnx into ref8
xadupre Jun 15, 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
6 changes: 3 additions & 3 deletions onnx/reference/reference_evaluator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,10 +7,10 @@

import numpy as np

from onnx import load, numpy_helper
from onnx import load
from onnx.defs import onnx_opset_version
from onnx.onnx_pb import FunctionProto, GraphProto, ModelProto, NodeProto, TypeProto
from onnx.reference.op_run import OpRun, RuntimeContextError
from onnx.reference.op_run import OpRun, RuntimeContextError, to_array_extended
from onnx.reference.ops_optimized import optimized_operators


Expand Down Expand Up @@ -344,7 +344,7 @@ def _init(self) -> None:
self.rt_inits_ = {}
self.rt_nodes_ = []
for init in self.inits_:
self.rt_inits_[init.name] = numpy_helper.to_array(init) # type: ignore[union-attr,arg-type]
self.rt_inits_[init.name] = to_array_extended(init) # type: ignore[union-attr,arg-type]
run_params = {
"log": lambda pattern, *args: self._log(10, pattern, *args),
"opsets": self.opsets,
Expand Down
24 changes: 24 additions & 0 deletions onnx/test/reference_evaluator_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3367,6 +3367,30 @@ def test_quantize_linear_e4m3(self):
got = ref.run(None, {"X": data})
assert_allclose(expected, got[0])

def test_quantize_linear_e4m3_initializer(self):
X = make_tensor_value_info("X", TensorProto.FLOAT, [None])
Y = make_tensor_value_info("Y", TensorProto.FLOAT, [None])
model = make_model(
make_graph(
[
make_node("QuantizeLinear", ["X", "scale", "zero"], ["T"]),
make_node("DequantizeLinear", ["T", "scale"], ["Y"], axis=0),
],
"g",
[X],
[Y],
[
make_tensor("scale", TensorProto.FLOAT, [1], [2.0]),
make_tensor("zero", TensorProto.FLOAT8E4M3FN, [1], [0.0]),
],
)
)
ref = ReferenceEvaluator(model)
data = np.array([0, 1, 2, 1e5, 200], dtype=np.float32)
expected = np.array([0, 1, 2, 896, 192], dtype=np.float32)
got = ref.run(None, {"X": data})
assert_allclose(expected, got[0])

def test_quantize_linear_e5m2(self):
X = make_tensor_value_info("X", TensorProto.FLOAT, [None])
Y = make_tensor_value_info("Y", TensorProto.FLOAT, [None])
Expand Down