I'm running a pytorch + ray hyperparameter optimization, and the output to screen while the algorithm is running is:
== Status ==
Current time: 2022-09-20 16:10:16 (running for 00:11:50.27)
Memory usage on this node: 29.0/128.7 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/8 CPUs, 1.0/1 GPUs, 0.0/83.03 GiB heap, 0.0/9.31 GiB objects (0.0/1.0 accelerator_type:P2000)
Result logdir: /home/runs/train_fn_2022-09-20_15-58-22
Number of trials: 1/1 (1 RUNNING)
+-------------------+----------+-------------------+------------+-----------+------------------+--------------+
| Trial name | status | loc | c_hidden | dp_rate | dp_rate_linear | num_layers |
|-------------------+----------+-------------------+------------+-----------+------------------+--------------|
| train_fn_0fcfd440 | RUNNING | 172.18.0.2:356387 | 64 | 0.3 | 0.1 | 3 |
+-------------------+----------+-------------------+------------+-----------+------------------+--------------+
You can see it's saying I've requested 1 GPU - but then 0.0/1.0 accelerators is mentioned further on in the same line.
Can someone tell me how to tell if pytorch is actually implementing a GPU as it should be?