caffe gpu-util is low

[Yuankui Zhu]

Hi guys,  I am using caffe to train the Enet for segmantation/

But I found my gpu util is very low and power consumption is low.

When I am training with tensorflow  is the gpu ulti is always full and the sound is like flying.

Anyone got idea on that?>

[Yuankui Zhu]

And this is my solver .   target img is 2000*256*256*3     gt is 2000*256*256*1  

thx

 

net: "/home/ewen/megami/ENet/prototxts/enet_train_encoder_decoder.prototxt"  # Change this to the absolute path to your model file

test_initialization: true

test_iter: 1

test_interval: 10000000

base_lr: 0.00001                       

lr_policy: "multistep"

gamma: 0.1

stepvalue: 10000

stepvalue: 30000

stepvalue: 50000

display: 20

momentum: 0.9

max_iter: 70000

weight_decay: 0.0002

snapshot: 5000

snapshot_prefix: "/home/ewen/megami/ENet/weights/snapshots_decoder/enet"  # Change this to the absolute path to where you wish to output solver snapshots

solver_mode: GPU

solver_type: ADAM

momentum2: 0.999

 

[Jonathan R. Williford]

Try increasing the batch size.

Jonathan

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users+unsubscribe@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/7f75e0bc-1281-44e5-bcba-47f243a07d31%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Yuankui Zhu]

Thx for your reply.

But in my case, increasing the batch_size did increase the memory usage   ,  like in this case batch 8    6GB    ///  batch 16    11GB.

But the gpu-util remains at low level like 30-40%.

在 2017年12月18日星期一 UTC+8下午7:30:31，Jonathan R. Williford写道：

To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users...@googlegroups.com.

To post to this group, send email to caffe...@googlegroups.com.