Add a new parameter 'global batchsize' for the batch size multiplicat…

…ion for multi-gpu case

benchmark/python/gluon/benchmark_gluon.py

@@ -41,11 +41,14 @@
 parser.add_argument('--num-batches', type=int, default=10)
 parser.add_argument('--gpus', type=str, default='',
  help='ordinates of gpus to use, can be "0,1,2" or empty for cpu only.')
+parser.add_argument('--global-batchsize', type=bool, default=True,
+ help='for multi-gpu case, the batchsize will not be multiplied with GPU number if set true.') 
 parser.add_argument('--type', type=str, default='inference', choices=['all', 'training', 'inference'])
 
 opt = parser.parse_args()
 
 num_batches = opt.num_batches
+global_bs = opt.global_batchsize
 dry_run = 10 # use 10 iterations to warm up
 batch_inf = [1, 16, 32, 64, 128, 256]
 batch_train = [1, 2, 4, 8, 16, 32, 64, 126, 256]
@@ -143,25 +146,37 @@ def train(network, batch_size, ctx):
  logging.info('device: %s', devs)
  if runtype == 'inference' or runtype == 'all':
  if bs != 0:
- batch_sizes = bs * max(1, num_gpus)
+ if not global_bs:
+ batch_sizes = bs * max(1, num_gpus)
+ else:
+ batch_sizes = bs
  fwd_time = score(network, batch_sizes, devs)
  fps = (batch_sizes * num_batches)/fwd_time
  logging.info(network + ' inference perf for BS %d is %f img/s', bs, fps)
  else:
  for batch_size in batch_inf:
- batch_sizes = batch_size * max(1, num_gpus)
+ if not global_bs:
+ batch_sizes = batch_size * max(1, num_gpus)
+ else:
+ batch_sizes = batch_size
  fwd_time = score(network, batch_sizes, devs)
  fps = (batch_sizes * num_batches) / fwd_time
  logging.info(network + ' inference perf for BS %d is %f img/s', batch_size, fps)
  if runtype == 'training' or runtype == 'all':
  if bs != 0:
- batch_sizes = bs * max(1, num_gpus)
+ if not global_bs:
+ batch_sizes = bs * max(1, num_gpus)
+ else:
+ batch_sizes = bs
  bwd_time = train(network, batch_sizes, devs)
  fps = (batch_sizes * num_batches) / bwd_time
  logging.info(network + ' training perf for BS %d is %f img/s', bs, fps)
  else:
  for batch_size in batch_train:
- batch_sizes = batch_size * max(1, num_gpus)
+ if not global_bs:
+ batch_sizes = batch_size * max(1, num_gpus)
+ else:
+ batch_sizes = batch_size
  bwd_time = train(network, batch_sizes, devs)
  fps = (batch_sizes * num_batches) / bwd_time
  logging.info(network + ' training perf for BS %d is %f img/s', batch_size, fps)