change duration compute methods #49
Conversation
fluid/resnet50.py
Outdated
| @@ -195,9 +199,9 @@ def run_benchmark(model, args): | |||
| im_num += label.shape[0] | |||
|
|
|||
| duration = time.time() - start_time | |||
| im_num = im_num - args.skip_batch_num * args.batch_size | |||
| #im_num = im_num - args.skip_batch_num * args.batch_size | |||
There was a problem hiding this comment.
please delete the codes commented
fluid/resnet50.py
Outdated
| if iter < args.skip_batch_num: | ||
| iter += 1 | ||
| continue | ||
| if pass_id == 0: | ||
| start_time = time.time() |
There was a problem hiding this comment.
For all batches in pass 0, this line will be executed. Is this what you want?
fluid/resnet50.py
Outdated
| accuracy.reset(exe) | ||
| if iter == args.iterations: | ||
| break | ||
| iter = 0 |
There was a problem hiding this comment.
Is this modification right? What's the use of args.iterations ?
There was a problem hiding this comment.
I think it shoud be:
accuracy.reset(exe)
iter = 0
the args.iterations is used to limit the iterations in each pass, so it used in:
for batch_id, data in enumerate(train_reader()):
fluid/resnet50.py
Outdated
| @@ -198,7 +198,7 @@ def run_benchmark(model, args): | |||
| buf_size=5120), | |||
| batch_size=args.batch_size) | |||
|
|
|||
| place = core.CPUPlace() if args.device == 'CPU' else core.CUDAPlace(0) | |||
| place = core.CPUPlace() if args.device == 'CPU' else core.GPUPlace(0) | |||
There was a problem hiding this comment.
I remember CUDAPlace is used in the latest PaddlePaddle. Could you refer to the source codes to confirm this.
fluid/resnet50.py
Outdated
| im_num = 0 | ||
| start_time = time.time() | ||
| for pass_id in range(args.pass_num): |
There was a problem hiding this comment.
https://github.com/ccmteorljh/benchmark/blob/6e9b36b2cd318f308b80c984617f79f6eccdce5e/fluid/resnet50.py#L170 Maybe moving L170 before L213 is better.
1,、修改了duration的计算方法，原来的是统计了最后一个pass的时间，但是计算examples_per_sec时im_num缺失所有pass的图片总和，现在修改duration为所有pass的时间总和；
2、skip_batch_num这个参数没有起到作用，现在修改了