cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
ru2308hoi
Visitor
Visitor
119 Views
Registered: ‎04-14-2021

inconsistency between deploy_model.pb and quantize_eval_model.pb

After we quantized using the command : 

vai_q_tensorflow quantize \

--input_frozen_graph ./model.pb \

--input_nodes x_input\

--input_shapes ?,160,160,3 \

--output_nodes output \

--method 1 \

--input_fn graph_input_fn.calib_input \

--gpu 0 \

--calib_iter 50 \

We found deploy_model.pb and quantize_eval_model.pb were very different. 

deploy_model.pb looks normal but we are not supposed to use it for vitis1.3

all wights in quantize_eval_model.pb shows 0 ..... why ? 

 

deploy_model.pb

ru2308hoi_0-1618496750837.png

 

 

quantize_eval_model.pb

 

ru2308hoi_0-1618497130421.png

 

 

0 Kudos
0 Replies