Improvements to quantizer: Removed unused qType field, add reshape op

[pranav-prakash]

• Remove the unused qType field. This was always set to uint8 anyway, which was incorrect for int8 quantization.
• Fix case when quantized bias already present in quantizer map
• Support for Reshape op added, which is a passthrough (no-op in terms of quantization).

[yufenglee]

/azp run Linux CPU CI Pipeline,Linux CPU x64 NoContribops CI Pipeline,Linux GPU CI Pipeline,Linux GPU TensorRT CI Pipeline,MacOS CI Pipeline,MacOS NoContribops CI Pipeline,Windows CPU CI Pipeline,Windows GPU CI Pipeline,Windows GPU TensorRT CI Pipeline

[yufenglee]

/azp run orttraining-linux-ci-pipeline,orttraining-mac-ci-pipeline,orttraining-linux-gpu-ci-pipeline,centos7_cpu,Linux CPU Minimal Build E2E CI Pipeline,Linux Nuphar CI Pipeline,MacOS NoContribops CI Pipeline,Linux OpenVINO CI Pipeline,orttraining-distributed

[azure-pipelines]

Azure Pipelines successfully started running 9 pipeline(s).

[azure-pipelines]

Azure Pipelines successfully started running 9 pipeline(s).

[yufenglee]

@pranav-prakash , it fails some unit tests. Could you please fix them?

[pranav-prakash]

@yufenglee Fixed. As part of fixing that I saw that in quant_utils the type of qType was inconsistent: in all other places qType is of type onnx_proto.TensorProto, but in quantize_nparray it was expected to be of type QuantType. This meant that there was actually a latent bug in the existing pad.py because in line 42 we had called

quantize_nparray(quantized_input_value.qType, padding_constant_array, scale_value, zp_value)

but then in line 52 we call

pad_value_qnodes = self.quantizer._get_quantize_input_nodes(node, 2, quantized_input_value.qType, ...

and _get_quantize_input_nodes expects qType to be of type onnx_proto.TensorProto.

I've cleaned it up so everything is consistent in using ``onnx_proto.TensorProto`.

[yufenglee]

/azp run Linux CPU CI Pipeline,Linux CPU x64 NoContribops CI Pipeline,Linux GPU CI Pipeline,Linux GPU TensorRT CI Pipeline,MacOS CI Pipeline,MacOS NoContribops CI Pipeline,Windows CPU CI Pipeline,Windows GPU CI Pipeline,Windows GPU TensorRT CI Pipeline

[yufenglee]

/azp run orttraining-linux-ci-pipeline,orttraining-mac-ci-pipeline,orttraining-linux-gpu-ci-pipeline,centos7_cpu,Linux CPU Minimal Build E2E CI Pipeline,Linux Nuphar CI Pipeline,MacOS NoContribops CI Pipeline,Linux OpenVINO CI Pipeline,orttraining-distributed

[azure-pipelines]

Azure Pipelines successfully started running 9 pipeline(s).

[azure-pipelines]

Azure Pipelines successfully started running 9 pipeline(s).

[yufenglee]

/azp run orttraining-amd-gpu-ci-pipeline

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).