Quantize Bias for Conv/Gemm on Quantized Model #22889
Open
+269
−0
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
centwang
force-pushed
the
weicwang/bias_quantization
branch
from
60f58bc
to
4860244
Compare
skottmckay
reviewed
Nov 26, 2024
Comment on lines
+90
to
+91
| // Bias is quantized to int32. | ||
| ONNX_NAMESPACE::TypeProto int32_type_proto; |
There was a problem hiding this comment.
Does this data type work in general or would it potentially be EP specific. i.e. if the EP uses the quantized op is it universally expected that the bias input would be int32 if the data type is say 8-bit int for the quantized Conv/Gemm?
There was a problem hiding this comment.
In https://github.com/onnx/onnx/blob/main/docs/Operators.md#QLinearConv, T4 (for bias) is tensor(int32) only. Same from QGemm schema, so I guess this work in general?
| // Bias DQ node produces output to Conv/Gemm node's input_2, with scale = input_scale_0 * input_scale_1, zp = 0. | ||
| NodeArg& bias_dq_node_arg = | ||
| graph.GetOrCreateNodeArg(graph.GenerateNodeArgName(node.Name() + "_bias_dq"), &bias_dq_type); | ||
| Node& dp_node = graph.AddNode(graph.GenerateNodeName(node.Name() + "_bias_dq"), QDQ::DQOpName, "Bias DQ node", |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Some quantized models don't have Conv/Gemm node's bias quantized but still leave them in float. This PR is to create a sub-graph to quantize the bias for Conv/Gemm nodes with scale = scale_input_0 * scale_input_1 and zp = 0. We only do this for bias initializer so that ConstantFolding will fold the sub-graph to a real quantized int32 bias initializer during the graph optimization next round.