![]() ![]() However, if you don't call get_available_gpus(), the memory allocation works fine. If you run the above code, you could notice that even though you set GPU memory fraction per process to 0.1, it still allocates the whole GPU memory by looking at command nvidia-smi. Gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.1)ĭevice_count=, Log.info("On this machine, GPU devices: ", get_available_gpus()) Local_device_protos = device_lib.list_local_devices() This would be compatible with both 32 bit and 64 bit windows.From import device_lib This is a full offline installer standalone setup for Windows Operating System.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |