out of VRAM about interactive-deep-colorization

[clickcell]

It worked for me in CPU mode, RAM usage was about 4.7GB, so I think it's quite far from fitting in the 3GB envelope, which is a shame, because performance in CPU mode was very poor for me. Now if only my laptop was the new model with 6GB VRAM, I imagine this would be running great.

[clickcell]

I reckon so. Let me know.

[ilkeraksoy]

I have similiar problem. I can't find batch_size. Please help me.

[ilkeraksoy]

How did you work in CPU mode? Thanks.

[clickcell]

I followed the instructions in the comment from JS1982 #4 (comment)

[ilkeraksoy]

Thanks! I will try...

[SleepProgger]

It worked for me when i reduced Xd / the load_size` parameter.
I am using 128 for ideepcolor and 192 for the coloring by reference.
You need to update the input layer in all prototxt files to match that value.
There is probably a nicer way (reduce the batch size), but i am new to caffe, so no clue about that.

[whizzzkid]

CPU mode seems to work, run it like python ideepcolor.py --gpu -1

-1 == cpu mode

[ilkeraksoy]

I tried it. But the same problem continues.

(NN_STATUS_SUCCESS (4 vs. 0) CUDNN_STATUS_INTERNAL_ERROR
*** Check failure stack trace: ***
Aborted (core dumped) )

[Boggartfly]

@clickcell My Titan X should work right? It has 12GB of VRAM.

[junyanz]

@ilkeryasinaksoy, could you report the full trace?