cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
shuakang
Visitor
Visitor
330 Views
Registered: ‎02-18-2020

An error when use decent_q tool to compress ssd-resnet34

Jump to solution

I have an error when I use decent_q tool to compress ssd-resnet-34:

tensorflow.python.framework.errors_impl.UnimplementedError: Generic conv implem entation does not support grouped convolutions for now.
[[Node: ssd1200/conv1/conv1_bn1/FusedBatchNorm/Mul = Conv2D[T=DT_FLOAT , data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true,_device="/job:localhost/replica:0/task:0/device:CPU :0"](_arg_image_0_0,ssd1200/conv1/conv1_1/Conv2D/merged_input)]]

 

Enviroment:

xilinx_DNNDK V3.0 for Xilinx ZCU102

Ubuntu 16.04 LTS

CPU Only

Python 3.6

tensorflow-1.9.0 (cpu version)

ssd-resnet-34 model file is come from Vitis-AI. Location is "Vitis-AI\AI-Model-Zoo\models\tf_mlperf_resnet34_coco_1200_1200_433G\float\ resnet34_tf.22.1.nhwc.pb"

0 Kudos
Reply
1 Solution

Accepted Solutions
graces
Moderator
Moderator
255 Views
Registered: ‎07-16-2008

Would you please try out the quantizer in Vitis AI? 

-------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution.
-------------------------------------------------------------------------

View solution in original post

0 Kudos
Reply
2 Replies
graces
Moderator
Moderator
256 Views
Registered: ‎07-16-2008

Would you please try out the quantizer in Vitis AI? 

-------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution.
-------------------------------------------------------------------------

View solution in original post

0 Kudos
Reply
shuakang
Visitor
Visitor
248 Views
Registered: ‎02-18-2020

I tried to use quantizer in Vitis AI to compress ssd-resnet34. This error disappare. But one other error occurred.

INFO: Calibration Done.
INFO: Generating Deploy Model...
2020-01-17 09:21:30.289903: F tensorflow/contrib/decent_q/utils/deploy_quantized_graph.cc:961] Check failed: param_node->op() == "Const" [DEPLOY ERROR] Expect const node for nodemap_1/while/select_bboxes/sort_bboxes_80/Pad_4(Pad)'s param, but ExpandDims(map_1/while/select_bboxes/sort_bboxes_80/ExpandDims) found.

Please see: 

https://forums.xilinx.com/t5/AI-and-Vitis-AI/Why-one-of-Pad-node-s-input-only-must-be-a-constant-node-when/m-p/1079437

Thanks!

0 Kudos
Reply