PyTorch 中通道在最后的内存格式(beta)
原文:https://pytorch.org/tutorials/intermediate/memory_format_tutorial.html
作者: Vitaly Fedyunin
什么是通道在最后
通道在最后的内存格式是在保留内存尺寸的顺序中对 NCHW 张量进行排序的另一种方法。 通道最后一个张量的排序方式使通道成为最密集的维度(又称为每像素存储图像)。
例如,NCHW 张量的经典(连续)存储(在我们的示例中是具有 3 个颜色通道的两个2x2
图像)如下所示:
通道最后的存储格式对数据的排序方式不同:
Pytorch 通过使用现有的跨步结构支持内存格式(并提供与现有模型(包括 eager,JIT 和 TorchScript)的向后兼容性)。 例如,通道在最后的格式中的10x3x16x16
批量的步幅等于(768, 1, 48, 3)
。
通道最后一个存储格式仅适用于 4D NCWH 张量。
import torch
N, C, H, W = 10, 3, 32, 32
内存格式 API
这是在连续和通道最后存储格式之间转换张量的方法。
经典 PyTorch 连续张量
x = torch.empty(N, C, H, W)
print(x.stride()) # Ouputs: (3072, 1024, 32, 1)
出:
(3072, 1024, 32, 1)
转换运算符
x = x.contiguous(memory_format=torch.channels_last)
print(x.shape) # Outputs: (10, 3, 32, 32) as dimensions order preserved
print(x.stride()) # Outputs: (3072, 1, 96, 3)
出:
torch.Size([10, 3, 32, 32])
(3072, 1, 96, 3)
返回连续
x = x.contiguous(memory_format=torch.contiguous_format)
print(x.stride()) # Outputs: (3072, 1024, 32, 1)
出:
(3072, 1024, 32, 1)
替代选择
x = x.to(memory_format=torch.channels_last)
print(x.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
格式检查
print(x.is_contiguous(memory_format=torch.channels_last)) # Ouputs: True
出:
True
最后创建为渠道
x = torch.empty(N, C, H, W, memory_format=torch.channels_last)
print(x.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
clone
保留内存格式
y = x.clone()
print(y.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
to
,cuda
,float
…保留内存格式
if torch.cuda.is_available():
y = x.cuda()
print(y.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
empty_like
和*_like
运算符保留内存格式
y = torch.empty_like(x)
print(y.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
点向运算符保留内存格式
z = x + y
print(z.stride()) # Ouputs: (3072, 1, 96, 3)
出:
(3072, 1, 96, 3)
转换,Batchnorm
模块支持通道在最后(仅适用于CudNN >= 7.6
)
if torch.backends.cudnn.version() >= 7603:
input = torch.randint(1, 10, (2, 8, 4, 4), dtype=torch.float32, device="cuda", requires_grad=True)
model = torch.nn.Conv2d(8, 4, 3).cuda().float()
input = input.contiguous(memory_format=torch.channels_last)
model = model.to(memory_format=torch.channels_last) # Module parameters need to be Channels Last
out = model(input)
print(out.is_contiguous(memory_format=torch.channels_last)) # Ouputs: True
出:
True
性能提升
在具有张量核心支持的 Nvidia 硬件上观察到了最大的性能提升。 在运行 Nvidia 提供的 AMP(自动混合精度)训练脚本时,我们可以将性能提高 22% 以上。
python main_amp.py -a resnet50 --b 200 --workers 16 --opt-level O2 ./data
# opt_level = O2
# keep_batchnorm_fp32 = None <class 'NoneType'>
# loss_scale = None <class 'NoneType'>
# CUDNN VERSION: 7603
# => creating model 'resnet50'
# Selected optimization level O2: FP16 training with FP32 batchnorm and FP32 master weights.
# Defaults for this optimization level are:
# enabled : True
# opt_level : O2
# cast_model_type : torch.float16
# patch_torch_functions : False
# keep_batchnorm_fp32 : True
# master_weights : True
# loss_scale : dynamic
# Processing user overrides (additional kwargs that are not None)...
# After processing overrides, optimization options are:
# enabled : True
# opt_level : O2
# cast_model_type : torch.float16
# patch_torch_functions : False
# keep_batchnorm_fp32 : True
# master_weights : True
# loss_scale : dynamic
# Epoch: [0][10/125] Time 0.866 (0.866) Speed 230.949 (230.949) Loss 0.6735125184 (0.6735) Prec@1 61.000 (61.000) Prec@5 100.000 (100.000)
# Epoch: [0][20/125] Time 0.259 (0.562) Speed 773.481 (355.693) Loss 0.6968704462 (0.6852) Prec@1 55.000 (58.000) Prec@5 100.000 (100.000)
# Epoch: [0][30/125] Time 0.258 (0.461) Speed 775.089 (433.965) Loss 0.7877287269 (0.7194) Prec@1 51.500 (55.833) Prec@5 100.000 (100.000)
# Epoch: [0][40/125] Time 0.259 (0.410) Speed 771.710 (487.281) Loss 0.8285319805 (0.7467) Prec@1 48.500 (54.000) Prec@5 100.000 (100.000)
# Epoch: [0][50/125] Time 0.260 (0.380) Speed 770.090 (525.908) Loss 0.7370464802 (0.7447) Prec@1 56.500 (54.500) Prec@5 100.000 (100.000)
# Epoch: [0][60/125] Time 0.258 (0.360) Speed 775.623 (555.728) Loss 0.7592862844 (0.7472) Prec@1 51.000 (53.917) Prec@5 100.000 (100.000)
# Epoch: [0][70/125] Time 0.258 (0.345) Speed 774.746 (579.115) Loss 1.9698858261 (0.9218) Prec@1 49.500 (53.286) Prec@5 100.000 (100.000)
# Epoch: [0][80/125] Time 0.260 (0.335) Speed 770.324 (597.659) Loss 2.2505953312 (1.0879) Prec@1 50.500 (52.938) Prec@5 100.000 (100.000)
传递--channels-last true
允许以通道在最后的格式运行模型,观察到 22% 的表现增益。
python main_amp.py -a resnet50 --b 200 --workers 16 --opt-level O2 --channels-last true ./data
# opt_level = O2
# keep_batchnorm_fp32 = None <class 'NoneType'>
# loss_scale = None <class 'NoneType'>
#
# CUDNN VERSION: 7603
#
# => creating model 'resnet50'
# Selected optimization level O2: FP16 training with FP32 batchnorm and FP32 master weights.
#
# Defaults for this optimization level are:
# enabled : True
# opt_level : O2
# cast_model_type : torch.float16
# patch_torch_functions : False
# keep_batchnorm_fp32 : True
# master_weights : True
# loss_scale : dynamic
# Processing user overrides (additional kwargs that are not None)...
# After processing overrides, optimization options are:
# enabled : True
# opt_level : O2
# cast_model_type : torch.float16
# patch_torch_functions : False
# keep_batchnorm_fp32 : True
# master_weights : True
# loss_scale : dynamic
#
# Epoch: [0][10/125] Time 0.767 (0.767) Speed 260.785 (260.785) Loss 0.7579724789 (0.7580) Prec@1 53.500 (53.500) Prec@5 100.000 (100.000)
# Epoch: [0][20/125] Time 0.198 (0.482) Speed 1012.135 (414.716) Loss 0.7007197738 (0.7293) Prec@1 49.000 (51.250) Prec@5 100.000 (100.000)
# Epoch: [0][30/125] Time 0.198 (0.387) Speed 1010.977 (516.198) Loss 0.7113101482 (0.7233) Prec@1 55.500 (52.667) Prec@5 100.000 (100.000)
# Epoch: [0][40/125] Time 0.197 (0.340) Speed 1013.023 (588.333) Loss 0.8943189979 (0.7661) Prec@1 54.000 (53.000) Prec@5 100.000 (100.000)
# Epoch: [0][50/125] Time 0.198 (0.312) Speed 1010.541 (641.977) Loss 1.7113249302 (0.9551) Prec@1 51.000 (52.600) Prec@5 100.000 (100.000)
# Epoch: [0][60/125] Time 0.198 (0.293) Speed 1011.163 (683.574) Loss 5.8537774086 (1.7716) Prec@1 50.500 (52.250) Prec@5 100.000 (100.000)
# Epoch: [0][70/125] Time 0.198 (0.279) Speed 1011.453 (716.767) Loss 5.7595844269 (2.3413) Prec@1 46.500 (51.429) Prec@5 100.000 (100.000)
# Epoch: [0][80/125] Time 0.198 (0.269) Speed 1011.827 (743.883) Loss 2.8196096420 (2.4011) Prec@1 47.500 (50.938) Prec@5 100.000 (100.000)
以下模型列表完全支持通道在最后,并在 Volta 设备上显示了 8%-35% 的表现增益:alexnet
,mnasnet0_5
,mnasnet0_75
,mnasnet1_0
,mnasnet1_3
,mobilenet_v2
,resnet101
,resnet152
,resnet18
,resnet34
,resnet50
,resnext50_32x4d
,shufflenet_v2_x0_5
,shufflenet_v2_x1_0
,shufflenet_v2_x1_5
,shufflenet_v2_x2_0
,squeezenet1_0
,squeezenet1_1
,vgg11
,vgg11_bn
,vgg13
,vgg13_bn
,vgg16
,vgg16_bn
,vgg19
,vgg19_bn
,wide_resnet101_2
,wide_resnet50_2
转换现有模型
通道在最后支持不受现有模型的限制,因为只要输入格式正确,任何模型都可以转换为通道在最后,并通过图传播格式。
# Need to be done once, after model initialization (or load)
model = model.to(memory_format=torch.channels_last) # Replace with your model
# Need to be done for every input
input = input.to(memory_format=torch.channels_last) # Replace with your input
output = model(input)
但是,并非所有运算符都完全转换为支持通道在最后(通常返回连续输出)。 这意味着您需要根据支持的运算符列表来验证已使用运算符的列表,或将内存格式检查引入急切的执行模式并运行模型。
运行以下代码后,如果运算符的输出与输入的存储格式不匹配,运算符将引发异常。
def contains_cl(args):
for t in args:
if isinstance(t, torch.Tensor):
if t.is_contiguous(memory_format=torch.channels_last) and not t.is_contiguous():
return True
elif isinstance(t, list) or isinstance(t, tuple):
if contains_cl(list(t)):
return True
return False
def print_inputs(args, indent=''):
for t in args:
if isinstance(t, torch.Tensor):
print(indent, t.stride(), t.shape, t.device, t.dtype)
elif isinstance(t, list) or isinstance(t, tuple):
print(indent, type(t))
print_inputs(list(t), indent=indent + ' ')
else:
print(indent, t)
def check_wrapper(fn):
name = fn.__name__
def check_cl(*args, **kwargs):
was_cl = contains_cl(args)
try:
result = fn(*args, **kwargs)
except Exception as e:
print("`{}` inputs are:".format(name))
print_inputs(args)
print('-------------------')
raise e
failed = False
if was_cl:
if isinstance(result, torch.Tensor):
if result.dim() == 4 and not result.is_contiguous(memory_format=torch.channels_last):
print("`{}` got channels_last input, but output is not channels_last:".format(name),
result.shape, result.stride(), result.device, result.dtype)
failed = True
if failed and True:
print("`{}` inputs are:".format(name))
print_inputs(args)
raise Exception(
'Operator `{}` lost channels_last property'.format(name))
return result
return check_cl
old_attrs = dict()
def attribute(m):
old_attrs[m] = dict()
for i in dir(m):
e = getattr(m, i)
exclude_functions = ['is_cuda', 'has_names', 'numel',
'stride', 'Tensor', 'is_contiguous', '__class__']
if i not in exclude_functions and not i.startswith('_') and '__call__' in dir(e):
try:
old_attrs[m][i] = e
setattr(m, i, check_wrapper(e))
except Exception as e:
print(i)
print(e)
attribute(torch.Tensor)
attribute(torch.nn.functional)
attribute(torch)
出:
Optional
'_Optional' object has no attribute '__name__'
如果您发现不支持通道在最后的张量的运算符并且想要贡献力量,请随时使用以下开发人员指南。
下面的代码是恢复火炬的属性。
for (m, attrs) in old_attrs.items():
for (k,v) in attrs.items():
setattr(m, k, v)
要做的工作
仍有许多事情要做,例如:
- 解决 N1HW 和 NC11 张量的歧义;
- 测试分布式训练支持;
- 提高运算符覆盖率。
如果您有反馈和/或改进建议,请通过创建 ISSUE 来通知我们。
脚本的总运行时间:(0 分钟 2.300 秒)
下载 Python 源码:memory_format_tutorial.py
下载 Jupyter 笔记本:memory_format_tutorial.ipynb
由 Sphinx 画廊生成的画廊