saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-06-06 00:03:22 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Allow Tensor to be scalar if it is not per channel. (#13959)

Browse source 
### Description
Allow Tensor to be scalar if it is not per channel.



### Motivation and Context
[<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here.
-->](https://github.com/microsoft/onnxruntime/issues/13915)
This commit is contained in:
Jian Chen 2022-12-14 13:23:56 -08:00 committed by GitHub
parent 5b492cbae3
commit e5f6689ae7
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 12 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
onnxruntime/python/tools/quantization/onnx_quantizer.py
View file

@ -691,13 +691,23 @@ class ONNXQuantizer:
# update scale initializer
quantized_bias_scale_name = quantized_bias_name + "_scale"
bias_scale_data = np.asarray(bias_scale, dtype=np.float32).reshape(-1)
packed_bias_scale_initializer = onnx.numpy_helper.from_array(bias_scale_data, quantized_bias_scale_name)
if self.is_per_channel():
packed_bias_scale_initializer = onnx.numpy_helper.from_array(bias_scale_data, quantized_bias_scale_name)
else:
packed_bias_scale_initializer = onnx.helper.make_tensor(
quantized_bias_scale_name, onnx_proto.TensorProto.FLOAT, [], bias_scale_data
)
self.model.initializer().extend([packed_bias_scale_initializer])
# update zero initializer
quantized_bias_zp_name = quantized_bias_name + "_zero_point"
bias_zp_data = np.zeros(bias_scale.shape, dtype=np.int32).reshape(-1)
packed_bias_zp_initializer = onnx.numpy_helper.from_array(bias_zp_data, quantized_bias_zp_name)
if self.is_per_channel():
packed_bias_zp_initializer = onnx.numpy_helper.from_array(bias_zp_data, quantized_bias_zp_name)
else:
packed_bias_zp_initializer = onnx.helper.make_tensor(
quantized_bias_zp_name, onnx_proto.TensorProto.INT32, [], bias_zp_data
)
self.model.initializer().extend([packed_bias_zp_initializer])
assert bias_name not in self.quantized_value_map