Forums - ONNX conversion crypt errors

1 post / 0 new

or Register

ONNX conversion crypt errors

rod

Join Date: 24 Jan 20

Posts: 1

Posted: Thu, 2020-02-27 15:01

Top

Hello,

I'm trying to convert a mobilenet v2 model that's been converted from torch to onnx.

This model was specifically choosen because it allow us to train in pytorch and convert to onnx using SNPE's supported opeset, however the conversion to DLC ends up in a undocumented error:

...
2020-02-27 19:51:26,688 - 145 - DEBUG_1 - Added buffer named 593 of shape [1, 576, 19, 19]
2020-02-27 19:51:26,688 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 115 with type Conv
2020-02-27 19:51:26,688 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,688 - 138 - DEBUG - Added OpNode with name convolution_40, in_names ['593'], out_names ['594']
2020-02-27 19:51:26,689 - 145 - DEBUG_1 - Added buffer named 594 of shape [1, 576, 10, 10]
2020-02-27 19:51:26,689 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 116 with type BatchNormalization
2020-02-27 19:51:26,689 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,689 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,689 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,689 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,690 - 138 - DEBUG - Added OpNode with name batchnorm_40, in_names ['594'], out_names ['595']
2020-02-27 19:51:26,690 - 145 - DEBUG_1 - Added buffer named 595 of shape [1, 576, 10, 10]
2020-02-27 19:51:26,690 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 117 with type Relu
2020-02-27 19:51:26,690 - 138 - DEBUG - Added OpNode with name neuron_27, in_names ['595'], out_names ['596']
2020-02-27 19:51:26,690 - 145 - DEBUG_1 - Added buffer named 596 of shape [1, 576, 10, 10]
2020-02-27 19:51:26,690 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 118 with type Conv
2020-02-27 19:51:26,690 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,691 - 138 - DEBUG - Added OpNode with name convolution_41, in_names ['596'], out_names ['597']
2020-02-27 19:51:26,691 - 145 - DEBUG_1 - Added buffer named 597 of shape [1, 160, 10, 10]
2020-02-27 19:51:26,691 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 119 with type BatchNormalization
2020-02-27 19:51:26,692 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,692 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,692 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,692 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,692 - 138 - DEBUG - Added OpNode with name batchnorm_41, in_names ['597'], out_names ['598']
2020-02-27 19:51:26,692 - 145 - DEBUG_1 - Added buffer named 598 of shape [1, 160, 10, 10]
2020-02-27 19:51:26,692 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 120 with type Conv
2020-02-27 19:51:26,693 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,693 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,693 - 138 - DEBUG - Added OpNode with name convolution_42, in_names ['593'], out_names ['599']
2020-02-27 19:51:26,693 - 145 - DEBUG_1 - Added buffer named 599 of shape [1, 576, 19, 19]
2020-02-27 19:51:26,693 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 121 with type BatchNormalization
2020-02-27 19:51:26,694 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,694 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,694 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,694 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,694 - 138 - DEBUG - Added OpNode with name batchnorm_42, in_names ['599'], out_names ['600']
2020-02-27 19:51:26,694 - 145 - DEBUG_1 - Added buffer named 600 of shape [1, 576, 19, 19]
2020-02-27 19:51:26,695 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 122 with type Clip
2020-02-27 19:51:26,695 - 138 - DEBUG - Added OpNode with name neuron_28, in_names ['600'], out_names ['601']
2020-02-27 19:51:26,695 - 145 - DEBUG_1 - Added buffer named 601 of shape [1, 576, 19, 19]
2020-02-27 19:51:26,695 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 123 with type Conv
2020-02-27 19:51:26,695 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,696 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,696 - 138 - DEBUG - Added OpNode with name convolution_43, in_names ['601'], out_names ['602']
2020-02-27 19:51:26,696 - 145 - DEBUG_1 - Added buffer named 602 of shape [1, 126, 19, 19]
2020-02-27 19:51:26,696 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 124 with type Transpose
2020-02-27 19:51:26,697 - 138 - DEBUG - Added OpNode with name permute_0, in_names ['602'], out_names ['603']
2020-02-27 19:51:26,697 - 145 - DEBUG_1 - Added buffer named 603 of shape [1, 19, 19, 126]
2020-02-27 19:51:26,697 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 125 with type Shape
2020-02-27 19:51:26,697 - 174 - WARNING - WARNING_STATIC_SHAPE: SNPE only supports static shape ops, interpreted at conversion time
2020-02-27 19:51:26,697 - 138 - DEBUG - Inserting weights for 604
2020-02-27 19:51:26,697 - 138 - DEBUG - Added OpNode with name static_0, in_names [], out_names []
2020-02-27 19:51:26,697 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 126 with type Slice
2020-02-27 19:51:26,698 - 138 - DEBUG - DEBUG_RETRIEVE_WEIGHTS, key:
2020-02-27 19:51:26,698 - 138 - DEBUG - Inserting weights for 605
2020-02-27 19:51:26,698 - 138 - DEBUG - Added OpNode with name static_1, in_names [], out_names []
2020-02-27 19:51:26,698 - 138 - DEBUG - DEBUG_CONVERTING_NODE: Attempting to convert node 127 with type Squeeze
2020-02-27 19:51:26,698 - 164 - ERROR - Node : '605'

Source model was taken from https://github.com/qfgaohao/pytorch-ssd , then converted to onnx.

Forum vote up/down

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.

Sort By

Filter Results