When I quantize the yolo_tube.onnx model, the scale_ array tensor for bias was a scalar, so an exception was thrown

[happyAnger66-an]

TensorRT-Model-Optimizer/modelopt/onnx/quantization/qdq_utils.py

Line 666 in e35db17

if scale_array.shape and scale_array.shape[0] != weight_shape[axis]:

Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/modelopt/onnx/quantization/qdq_utils.py", line 781, in qdq_to_dq
scaled = _convert_weight(weight_array, scale, zp, quantized_node)
File "/usr/local/lib/python3.10/dist-packages/modelopt/onnx/quantization/qdq_utils.py", line 676, in _convert_weight
reshape_dims[axis] = scale_array.shape[0]
IndexError: tuple index out of range

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/srcs/codes/wujie/model-optimizer/./src/model_optimizer/quantization/models/yolo.py", line 114, in
quantize_model_onnx(model, quant_cfg, quant_mode, args)
File "/srcs/codes/wujie/model-optimizer/./src/model_optimizer/quantization/models/yolo.py", line 39, in quantize_model_onnx
quantize(
File "/usr/local/lib/python3.10/dist-packages/modelopt/onnx/quantization/quantize.py", line 523, in quantize
onnx_model = qdq_to_dq(onnx_model)
File "/usr/local/lib/python3.10/dist-packages/modelopt/onnx/quantization/qdq_utils.py", line 806, in qdq_to_dq
raise RuntimeError(f"Failed to convert node {node.name}: {e!s}")
RuntimeError: Failed to convert node /model.23/cv3.0/cv3.0.2/weight_quantizer/QuantizeLinear: tuple index out of range

After adding the following code, the problem can be fixed
scale_array = onnx.numpy_helper.to_array(scale)
if not scale_array.shape: ### ---------this is my added
scale_array = np.expand_dims(scale_array, axis=0)
print(f'none scale found extend to {scale_array}')

[happyAnger66-an]

After adding the following code, the problem can be fixed

scale_array = onnx.numpy_helper.to_array(scale)
if not scale_array.shape: ### ---------this is my added
scale_array = np.expand_dims(scale_array, axis=0)
print(f'none scale found extend to {scale_array}')