Danielyijun

Results 10 comments of Danielyijun

Could you help me to check it out? Really appreciate it. Thank you :) @sj6077

Yes, I followed all the guide to install TensorFlow and horovod. @sj6077

No, how can I do it? @sj6077

Oh. If I got it right, I should have chosen y for nGraph. But I will try to reinstall tensorflow and make sure it. I'll let you know how it...

May I ask does the bazel version matter? Do you have any recommendation as well?

Okay, I got it, thank you. :)

Hi, I disabled nGraph when installing tensorflow but the issue was still there. ---------------------------------------------------- 2020-05-14 09:24:36.503557: E tensorflow/core/framework/op_segment.cc:53] Create kernel failed: Not found: No registered 'NGraphVariable' OpKernel for GPU devices...

> Can you run any code without parallax? If the error occurs again, it's not the parallax issue. Hi, I tried this way finding that I have another tensorflow folder...

I tried to run another command in terminal: CUDA_VISIBLE_DEVICES='' python3 /tmp/parallax-jyi/launch_ps.py --job_name=ps --task_index=0 --protocol=grpc --ps_hosts=10.0.0.103:36311,10.0.0.108:45713 --worker_hosts=10.0.0.103:44472,10.0.0.103:45326,10.0.0.103:40853,10.0.0.103:45386,10.0.0.108:38017,10.0.0.108:45945,10.0.0.108:46772,10.0.0.108:38564 This makes the whole training start, I wondered why we need to run another...