提升 5

二，环境配置

0，检查mac型号

点击桌面左上角mac图标——>关于本机——>概览，确定是m1芯片，了解内存大小(最好有16G以上，8G可能不太够用)。

1，下载 miniforge3 (miniforge3可以理解成 miniconda/annoconda 的社区版，提供了更稳定的对M1芯片的支持)

https://github.com/conda-forge/miniforge/#download

备注: annoconda 在 2022年5月开始也发布了对 mac m1芯片的官方支持，但还是推荐社区发布的miniforge3，开源且更加稳定。

2，安装 miniforge3

chmod +x ~/Downloads/Miniforge3-MacOSX-arm64.sh

sh ~/Downloads/Miniforge3-MacOSX-arm64.sh

source~/miniforge3/bin/activate

3，安装 pytorch (v1.12版本已经正式支持了用于mac m1芯片gpu加速的mps后端。)

pip install torch>=1.12 -i https://pypi.tuna.tsinghua.edu.cn/simple

4，测试环境

importtorch

print(torch.backends.mps.is_available)

print(torch.backends.mps.is_built)

如果输出都是True的话，那么恭喜你配置成功了。

三，范例代码

核心操作非常简单，和使用cuda类似，训练前把模型和数据都移动到torch.device("mps")就可以了。

importtorch

fromtorch importnn

importtorchvision

fromtorchvision importtransforms

importtorch.nn.functional asF

importos,sys,time

importnumpy asnp

importpandas aspd

importdatetime

fromtqdm importtqdm

fromcopy importdeepcopy

fromtorchmetrics importAccuracy

defprintlog(info):

nowtime = datetime.datetime.now.strftime( '%Y-%m-%d %H:%M:%S')

print( "\n"+ "=========="* 8+ "%s"%nowtime)

print(str(info)+ "\n")

#================================================================================

# 一，准备数据

#================================================================================

transform = transforms.Compose([transforms.ToTensor])

ds_train = torchvision.datasets.MNIST(root= "mnist/",train= True,download= True,transform=transform)

ds_val = torchvision.datasets.MNIST(root= "mnist/",train= False,download= True,transform=transform)

dl_train = torch.utils.data.DataLoader(ds_train, batch_size= 128, shuffle= True, num_workers= 2)

dl_val = torch.utils.data.DataLoader(ds_val, batch_size= 128, shuffle= False, num_workers= 2)

#================================================================================

# 二，定义模型

#================================================================================

defcreate_net:

net = nn.Sequential

net.add_module( "conv1",nn.Conv2d(in_channels= 1,out_channels= 64,kernel_size = 3))

net.add_module( "pool1",nn.MaxPool2d(kernel_size = 2,stride = 2))

net.add_module( "conv2",nn.Conv2d(in_channels= 64,out_channels= 512,kernel_size = 3))

net.add_module( "pool2",nn.MaxPool2d(kernel_size = 2,stride = 2))

net.add_module( "dropout",nn.Dropout2d(p = 0.1))

net.add_module( "adaptive_pool",nn.AdaptiveMaxPool2d(( 1, 1)))

net.add_module( "flatten",nn.Flatten)

net.add_module( "linear1",nn.Linear( 512, 1024))

net.add_module( "relu",nn.ReLU)

net.add_module( "linear2",nn.Linear( 1024, 10))

returnnet

net = create_net

print(net)

# 评估指标

classAccuracy(nn.Module):

def__init__(self):

super.__init__

self.correct = nn.Parameter(torch.tensor( 0.0),requires_grad= False)

self.total = nn.Parameter(torch.tensor( 0.0),requires_grad= False)

defforward(self, preds: torch.Tensor, targets: torch.Tensor):

preds = preds.argmax(dim= -1)

m = (preds == targets).sum

n = targets.shape[ 0]

self.correct += m

self.total += n

returnm/n

defcompute(self):

returnself.correct.float / self.total

defreset(self):

self.correct -= self.correct

self.total -= self.total

#================================================================================

# 三，训练模型

#================================================================================

loss_fn = nn.CrossEntropyLoss

optimizer= torch.optim.Adam(net.parameters,lr = 0.01)

metrics_dict = nn.ModuleDict({ "acc":Accuracy})

# =========================移动模型到mps上==============================

device = torch.device( "mps"iftorch.backends.mps.is_available else"cpu")

net.to(device)

loss_fn.to(device)

metrics_dict.to(device)

# ====================================================================

epochs = 20

ckpt_path= 'checkpoint.pt'

#early_stopping相关设置

monitor= "val_acc"

patience= 5

mode= "max"

history = {}

forepoch inrange( 1, epochs+ 1):

printlog( "Epoch {0} / {1}".format(epoch, epochs))

# 1，train -------------------------------------------------

net.train

total_loss,step = 0, 0

loop = tqdm(enumerate(dl_train), total =len(dl_train),ncols= 100)

train_metrics_dict = deepcopy(metrics_dict)

fori, batch inloop:

features,labels = batch

# =========================移动数据到mps上==============================

features = features.to(device)

labels = labels.to(device)

# ====================================================================

#forward

preds = net(features)

loss = loss_fn(preds,labels)

#backward

loss.backward

optimizer.step

optimizer.zero_grad

#metrics

step_metrics = { "train_"+name:metric_fn(preds, labels).item

forname,metric_fn intrain_metrics_dict.items}

step_log = dict({ "train_loss":loss.item},**step_metrics)

total_loss += loss.item

step+= 1

ifi!=len(dl_train) -1:

loop.set_postfix(**step_log)

else:

epoch_loss = total_loss/step

epoch_metrics = { "train_"+name:metric_fn.compute.item

forname,metric_fn intrain_metrics_dict.items}

epoch_log = dict({ "train_loss":epoch_loss},**epoch_metrics)

loop.set_postfix(**epoch_log)

forname,metric_fn intrain_metrics_dict.items:

metric_fn.reset

forname, metric inepoch_log.items:

history[name] = history.get(name, []) + [metric]

# 2，validate -------------------------------------------------

net.eval

total_loss,step = 0, 0

loop = tqdm(enumerate(dl_val), total =len(dl_val),ncols= 100)

val_metrics_dict = deepcopy(metrics_dict)

withtorch.no_grad:

fori, batch inloop:

features,labels = batch

# =========================移动数据到mps上==============================

features = features.to(device)

labels = labels.to(device)

# ====================================================================

#forward

preds = net(features)

loss = loss_fn(preds,labels)

#metrics

step_metrics = { "val_"+name:metric_fn(preds, labels).item

forname,metric_fn inval_metrics_dict.items}

step_log = dict({ "val_loss":loss.item},**step_metrics)

total_loss += loss.item

step+= 1

ifi!=len(dl_val) -1:

loop.set_postfix(**step_log)

else:

epoch_loss = (total_loss/step)

epoch_metrics = { "val_"+name:metric_fn.compute.item

forname,metric_fn inval_metrics_dict.items}

epoch_log = dict({ "val_loss":epoch_loss},**epoch_metrics)

loop.set_postfix(**epoch_log)

forname,metric_fn inval_metrics_dict.items:

metric_fn.reset

epoch_log[ "epoch"] = epoch

forname, metric inepoch_log.items:

history[name] = history.get(name, []) + [metric]

# 3，early-stopping -------------------------------------------------

arr_scores = history[monitor]

best_score_idx = np.argmax(arr_scores) ifmode== "max"elsenp.argmin(arr_scores)

ifbest_score_idx==len(arr_scores) -1:

torch.save(net.state_dict,ckpt_path)

print( ">".format(monitor,

arr_scores[best_score_idx]),file=sys.stderr)

iflen(arr_scores)-best_score_idx>patience:

print( ">".format(

monitor,patience),file=sys.stderr)

break

net.load_state_dict(torch.load(ckpt_path))

dfhistory = pd.DataFrame(history)

四，使用torchkeras支持Mac M1芯片加速

我在最新的3.3.0的torchkeras版本中引入了对 mac m1芯片的支持，当存在可用的 mac m1芯片/ GPU 时，会默认使用它们进行加速，无需做任何配置。

使用范例如下。😋😋😋

!pip install torchkeras>= 3.3.0

importnumpy asnp

importpandas aspd

frommatplotlib importpyplot asplt

importtorch

fromtorch importnn

importtorch.nn.functional asF

fromtorch.utils.data importDataset,DataLoader

importtorchkeras #Attention this line

#================================================================================

# 一，准备数据

#================================================================================

importtorchvision

fromtorchvision importtransforms

transform = transforms.Compose([transforms.ToTensor])

ds_train = torchvision.datasets.MNIST(root= "mnist/",train= True,download= True,transform=transform)

ds_val = torchvision.datasets.MNIST(root= "mnist/",train= False,download= True,transform=transform)

dl_train = torch.utils.data.DataLoader(ds_train, batch_size= 128, shuffle= True, num_workers= 2)

dl_val = torch.utils.data.DataLoader(ds_val, batch_size= 128, shuffle= False, num_workers= 2)

forfeatures,labels indl_train:

break

#================================================================================

# 二，定义模型

#================================================================================

defcreate_net:

net = nn.Sequential

net.add_module( "conv1",nn.Conv2d(in_channels= 1,out_channels= 64,kernel_size = 3))

net.add_module( "pool1",nn.MaxPool2d(kernel_size = 2,stride = 2))

net.add_module( "conv2",nn.Conv2d(in_channels= 64,out_channels= 512,kernel_size = 3))

net.add_module( "pool2",nn.MaxPool2d(kernel_size = 2,stride = 2))

net.add_module( "dropout",nn.Dropout2d(p = 0.1))

net.add_module( "adaptive_pool",nn.AdaptiveMaxPool2d(( 1, 1)))

net.add_module( "flatten",nn.Flatten)

net.add_module( "linear1",nn.Linear( 512, 1024))

net.add_module( "relu",nn.ReLU)

net.add_module( "linear2",nn.Linear( 1024, 10))

returnnet

net = create_net

print(net)

# 评估指标

classAccuracy(nn.Module):

def__init__(self):

super.__init__

self.correct = nn.Parameter(torch.tensor( 0.0),requires_grad= False)

self.total = nn.Parameter(torch.tensor( 0.0),requires_grad= False)

defforward(self, preds: torch.Tensor, targets: torch.Tensor):

preds = preds.argmax(dim= -1)

m = (preds == targets).sum

n = targets.shape[ 0]

self.correct += m

self.total += n

returnm/n

defcompute(self):

returnself.correct.float / self.total

defreset(self):

self.correct -= self.correct

self.total -= self.total

#================================================================================

# 三，训练模型

#================================================================================

model = torchkeras.KerasModel(net,

loss_fn = nn.CrossEntropyLoss,

optimizer= torch.optim.Adam(net.parameters,lr= 0.001),

metrics_dict = { "acc":Accuracy}

)

fromtorchkeras importsummary

summary(model,input_data=features);

# if gpu/mps is available, will auto use it, otherwise cpu will be used.

dfhistory=model.fit(train_data=dl_train,

val_data=dl_val,

epochs= 15,

patience= 5,

monitor= "val_acc",mode= "max",

ckpt_path= 'checkpoint.pt')

#================================================================================

# 四，评估模型

#================================================================================

model.evaluate(dl_val)

#================================================================================

# 五，使用模型

#================================================================================

model.predict(dl_val)[ 0: 10]

#================================================================================

# 六，保存模型

#================================================================================

# The best net parameters has been saved at ckpt_path='checkpoint.pt' during training.

net_clone = create_net

net_clone.load_state_dict(torch.load( "checkpoint.pt"))

五，M1芯片与CPU和Nvidia GPU速度对比

使用以上代码作为范例，分别在CPU, mac m1芯片，以及Nvidia GPU上 运行。

得到的运行速度截图如下：

纯CPU跑效果

Mac M1 芯片加速效果

Tesla P100 GPU加速效果

纯CPU跑一个epoch大约是3min 18s。

使用mac m1芯片加速，一个epoch大约是33 s，相比CPU跑，加速约6倍。

这和pytorch官网显示的训练过程平均加速7倍相当。

使用Nvidia Tesla P100 GPU加速，一个epoch大约是 8s，相比CPU跑，加速约25倍。

整体来说Mac M1芯片对 深度学习训练过程的加速还是非常显著的，通常达到5到7倍左右。

不过目前看和企业中最常使用的高端的Tesla P100 GPU相比，还是有2到4倍的训练速度差异，可以视做一个mini版的GPU吧。

因此Mac M1芯片比较适合本地训练一些中小规模的模型，快速迭代idea，使用起来还是蛮香的。

尤其是本来就打算想换个电脑的，用mac做开发本来比windows好使多了。

- EOF -

加主页君微信，不仅Python技能+1

主页君日常还会在个人微信分享 Python相关工具、资源和 精选技术文章，不定期分享一些 有意思的活动、 岗位内推以及 如何用技术做业余项目

加个微信，打开一扇窗

点击标题可跳转

1、 搞懂 Transformer 结构，看这篇 PyTorch 实现就够了

2、 冷门但好用的 Python 库推荐一波

3、 10个超级实用的数据可视化图表总结！

觉得本文对你有帮助？请分享给更多人

推荐关注「Python开发者」，提升Python技能

点赞和在看就是最大的支持❤️返回搜狐，查看更多