encichou

> Yes, that is right. hello,how to calculate on GPU,change cpu to gpu?

hello ,my cold run slowly,your code could run fast?

I use your code,but I only could process 64images per second using v100 gpu ，do you know what should I do?thanks for your reply

1. i have about 20 cpus 2. 3.RAM 300G 4.batchsize 64 5.my image shape about 256x256 .i need resize. my main trouble is that my gpu occupancy rate is too...

> 它不应该是0. > 你能分享一下你的命令行吗？  many parameters is your default parameters.

> 您可以将log_device_placement设置为True并共享设备放置信息吗？ ok, The computer is not around me，i will share my device placement info ,9 hours later.

hello,I have same question。do you find some function to disable hybrization? and i have another quesion that is how to release gpu memory? could you help me?

> GPU memory held by memory pool can be released by `mx.gpu(0).empty_cache()` thank you for your reply, i will try your suggestion.

> np. if it works, it would be great if you could help update the doc linked in the issue to mention this functionality. ok ,I am trying the function....

hello ,author, i meet the same question.could you tell me how to solve the warning