Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!31161 [MSLite][Micro]fix quantization model codegen for MCUs 

Merge pull request !31161 from lz/fix_quantization_micro
This commit is contained in:
i-robot 2022-03-11 12:33:00 +00:00 committed by Gitee
parent a1e426859f ce35cb021d
commit 55a32e433a
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
1 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
mindspore/lite/tools/converter/micro/coder/opcoders/base/quant_dtype_cast_coder.cc
View File

@ -41,9 +41,8 @@ int QuantDTypeCastCoder::DoCode(CoderContext *const context) {
MS_LOG(ERROR) << "QuantDTypeCast need quantization parameters which is not found.";
return RET_ERROR;
}
auto quant_arg = (!output_tensor_->quant_params().empty() && output_tensor_->quant_params().at(0).inited)
? output_tensor_->quant_params().at(0)
: input_tensor_->quant_params().at(0);
auto quant_arg = (output_tensor_->quant_params().empty()) ? input_tensor_->quant_params().at(0)
: output_tensor_->quant_params().at(0);
int num_unit_thread = input_tensor_->ElementsNum();
Collect(context,