search

lfads / lfads-run-manager

Matlab interface for Latent Factor Analysis via Dynamical Systems (LFADS)
https://lfads.github.io/lfads-run-manager
Apache License 2.0
50 stars 29 forks source link

gpu_query #12

Closed cindyhfls closed 5 years ago

cindyhfls commented 5 years ago

Hi I am trying the example Lorenz Experiment single session on Mac.

My code broke at query_gpu_status. Can anyone help with it??

Traceback (most recent call last): File "run_lfadsqueue.py", line 10, in tasks = lq.run_lfads_queue(queue_name, tensorboard_script, task_specs, gpu_list=gpu_list, one_task_per_gpu=True) File "/Users/jiaxintu/lfads-run-manager/src/lfadsqueue.py", line 432, in run_lfads_queue gpu_status = query_gpu_status() File "/Users/jiaxintu/lfads-run-manager/src/lfadsqueue.py", line 107, in query_gpu_status '--format=csv,noheader,nounits']) File "/Users/jiaxintu/anaconda3/envs/lfads/lib/python2.7/subprocess.py", line 216, in check_output process = Popen(stdout=PIPE, *popenargs, **kwargs) File "/Users/jiaxintu/anaconda3/envs/lfads/lib/python2.7/subprocess.py", line 394, in init errread, errwrite) File "/Users/jiaxintu/anaconda3/envs/lfads/lib/python2.7/subprocess.py", line 1047, in _execute_child raise child_exception OSError: [Errno 2] No such file or directory

djoshea commented 5 years ago

Ah, sorry, somehow I missed this notification. The core issue is that tensorflow GPU is not supported on Mac OS. In this specific instance, the call to nvidia-smi is failing, since it exists only on Linux installs of NVIDIA drivers. If you're trying to run on Mac with CPU, it will be slower, though it can be done. In this case you should just run the lfads_train.sh script directly rather than using the lfadsqueue.

cindyhfls commented 5 years ago

Thank you. I have figured it out too.