xinyu-intel commented on a change in pull request #10391: [MXNET-139] Tutorial
for mixed precision training with float16
URL: https://github.com/apache/incubator-mxnet/pull/10391#discussion_r198707423
##########
File path: example/image-classification/benchmark_score.py
##########
@@ -74,11 +75,17 @@ def score(network, dev, batch_size, num_batches):
devs.append(mx.cpu())
batch_sizes = [1, 2, 4, 8, 16, 32]
-
for net in networks:
logging.info('network: %s', net)
for d in devs:
logging.info('device: %s', d)
for b in batch_sizes:
- speed = score(network=net, dev=d, batch_size=b, num_batches=10)
- logging.info('batch size %2d, image/sec: %f', b, speed)
+ for dtype in ['float32', 'float16']:
+ if d == mx.cpu() and dtype == 'float16':
+ #float16 is not supported on CPU
+ continue
+ elif net in ['inception-bn', 'alexnet'] and dt ==
'float16':
Review comment:
Benchmark crash here since dt is not defined.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
