cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Highlighted
Participant
Participant
697 Views
Registered: ‎10-25-2018

DNNC_error

Jump to solution

Hi guys,

I quantized model mobilenet_ssd_v1_300x300 on VOC datasets. When I used decent_q that gave warning :

[DEPLOY WARNING] Node scale_logits(Type: RealDiv) is not quantized and cannot be deployed to DPU,because it has unquantized input node: scale_logits/y. Please deploy it on CPU.
[DEPLOY WARNING] Node convert_scores(Type: Sigmoid) is not quantized and cannot be deployed to DPU,because it has unquantized input node: scale_logits. Please deploy it on CPU.
[DEPLOY WARNING] Node raw_outputs/class_predictions(Type: Identity) is not quantized and cannot be deployed to DPU,because it has unquantized input node: convert_scores. Please deploy it on CPU.
INFO: Deploy Model Generated.

Therefore, dnnc_ZCU102.sh failed:

[DNNC][Error] 'Const' op should be fused with current op [RealDiv] by DECENT.

So what kind of this error, Could you help me to solve it ?

0 Kudos
1 Solution

Accepted Solutions
Highlighted
Participant
Participant
661 Views
Registered: ‎10-25-2018
0 Kudos
4 Replies
Highlighted
Participant
Participant
662 Views
Registered: ‎10-25-2018
0 Kudos
Highlighted
Contributor
Contributor
610 Views
Registered: ‎08-02-2019

Hi,

I am also facing the same kind of issue. Can you tell me how did you fix the error.

0 Kudos
Highlighted
Participant
Participant
526 Views
Registered: ‎10-25-2018

hi @jbond1007

In fact, I just skiped this node by --ignore_nodes flag and implement this node on CPU

0 Kudos
Highlighted
255 Views
Registered: ‎01-29-2020

I am facing same error. Can you please tell me how to implement unquantized nodes on CPU?

0 Kudos