Closed conghuawang closed 6 years ago
Sorry to hear what happened to you. CNTK maximizes GPU when training computation intensive models like ResNet, and thus it's an extreme test for hardware's thermal dissipation designs. If you look at nvidia-smi, you'll find GPU running at close to 100% when training ResNet20 CIFAR10. If this is a problem, please consider inserting some sleep in the training loop.
@KeDengMS Thanks for your reply. Today I bought a Alienware aurora to use CNTK .
Hello, I'm run "Examples\Image\Classification\ResNet\Python" in windows. But the example burn my motherboard twice!
In "Examples\Image\Classification\ResNet\Python\TrainResNet_CIFAR10.py", when this code is running, my motherboard burn twice: train_and_evaluate(reader_train, reader_test, network_name, epoch_size, epochs, args['profiler_dir'], model_dir, args['logdir'], args['tensorboard_logdir'], gen_heartbeat=args['genheartbeat'], fp16=args['fp16'])
My laptop is Alienware 17 R4 with GTX 1080, I can't run CNTK in laptop?
DxDiag infos below: DxDiag.txt