Forums - multiple out_node is not working on snpe-pytorch-to-dlc

2 posts / 0 new
Last post
multiple out_node is not working on snpe-pytorch-to-dlc
ckdtjr9775
Join Date: 8 Nov 20
Posts: 1
Posted: Tue, 2021-09-28 02:52

hi, i'm trying to convert SSD pytorch model to DLC using SNPE version 1.54.2.

my SSD model has two output node, "concatenation_0_0" for bbox and "concatenation_1_0" for class score

I can convert model with one of two output node, but I can't convert with both of them.

for example, 

snpe-pytorch-to-dlc --input_network ./ssd_300_traced.jit.pt -d x 1,3,300,300 --out_node cancatenation_0_0

produces cancatenation_0_0.raw and

snpe-pytorch-to-dlc --input_network ./ssd_300_traced.jit.pt -d x 1,3,300,300 --out_node concatenation_1_0 

produces concatenation_1_0 .raw

but,

snpe-pytorch-to-dlc --input_network ./ssd_300_traced.jit.pt -d x 1,3,300,300 --out_node concatenation_0_0 --out_node cancatenation_1_0 

produces only one output node, node cancatenation_1_0.raw

is there any solution for this issue?

 

 

 

 

  • Up0
  • Down0
weihuan
Join Date: 12 Apr 20
Posts: 270
Posted: Tue, 2021-11-23 18:15

Dear customer,

Thanks for your efforts in our products.

Please just remove the --out_node inside SNPE ONNX converter.  And then write the out_node with #output to your input list file while you run on final devices. Like as below.

#output1 output2

input:=1.raw

That may resolve your questions.

BR.

Wei

  • Up0
  • Down0
or Register

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.