Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.
Dear developer,
We support SNPE UDL rather than ONNX UDL since the conception of UDL is from SNPE inner not ONNX.
BR.
Wei
Hi,
Thanks for you reply. I think I need to clarify my question.
Currently we have a Caffe model with custom layers. When converting this caffe model to SNPE DLC, the custom layers were implemented using UDL. The snpe-caffe-to-dlc binary/script has an option to pass information about UDL.
We want to use a ONNX representation of the above model to convert to dlc (it is not exactly the same model but a similar one).
However the snpe-onnx-to-dlc script doesn't have an option to pass information about UDL. However I can provide UDO implementations of the custom layers if present. See the link below - one of the options to the tool is UDO config paths. My understanding is that if there are custom layers in ONNX, SNPE will support the UDO implementations of those custom layers
https://developer.qualcomm.com/sites/default/files/docs/snpe/tools.html#...
My question is whether UDL implementations were ever supported in snpe-onnx-to-dlc tool ? (for similarity see snpe-caffe-to-dlc tool which has an option to pass information about UDL).
Thanks.
Hi,
UDO is supported from SNPE-1.35 for all the frameworks, as mentioned in the SNPE document Tools UDL is supported in caffe SNPE-1.24 and not available for ONNX.
Please check the below mentioned link for the updates from SNPE release notes.
https://developer.qualcomm.com/software/qualcomm-neural-processing-sdk/r...
Thanks.