Weights & Biases (wandb): Difference between revisions

Jump to navigation Jump to search

Weights & Biases (wandb) (view source)

Revision as of 18:14, 8 February 2021

459 bytes added ,  3 years ago
Marked this version for translation
No edit summary
(Marked this version for translation)
Line 1: Line 1:
<languages />
<languages />
<translate>
<translate>
<!--T:1-->
[https://wandb.ai Weights & Biases (wandb)] is a "meta machine learning platform" designed to help AI practitioners and teams build reliable machine learning models for real-world applications by streamlining the machine learning model lifecycle. By using wandb, users can track, compare, explain and reproduce their machine learning experiments.
[https://wandb.ai Weights & Biases (wandb)] is a "meta machine learning platform" designed to help AI practitioners and teams build reliable machine learning models for real-world applications by streamlining the machine learning model lifecycle. By using wandb, users can track, compare, explain and reproduce their machine learning experiments.


== Using wandb on Compute Canada clusters ==
== Using wandb on Compute Canada clusters == <!--T:2-->


=== Availability ===  
=== Availability === <!--T:3-->




<!--T:4-->
Since it requires an internet connection, wandb has restricted availability on compute nodes, depending on the cluster:
Since it requires an internet connection, wandb has restricted availability on compute nodes, depending on the cluster:




<!--T:5-->
{| class="wikitable"
{| class="wikitable"
|-
|-
Line 22: Line 25:
|}
|}


=== Example ===
=== Example === <!--T:6-->


<!--T:7-->
The following is an example of how to use wandb to track experiments on Béluga. To reproduce this on Cedar, it is not necessary to load the module <tt>httpproxy</tt>.
The following is an example of how to use wandb to track experiments on Béluga. To reproduce this on Cedar, it is not necessary to load the module <tt>httpproxy</tt>.


<!--T:8-->
{{File
{{File
   |name=wandb-test.sh
   |name=wandb-test.sh
Line 37: Line 42:




<!--T:9-->
module load python/3.6 httpproxy
module load python/3.6 httpproxy
virtualenv --no-download $SLURM_TMPDIR/env
virtualenv --no-download $SLURM_TMPDIR/env
Line 43: Line 49:




<!--T:10-->
### Save your wandb API key in your .bash_profile or replace $API_KEY with your actual API key:  
### Save your wandb API key in your .bash_profile or replace $API_KEY with your actual API key:  


<!--T:11-->
wandb login $API_KEY  
wandb login $API_KEY  


<!--T:12-->
python wandb-test.py
python wandb-test.py
}}
}}


<!--T:13-->
The script wandb-test.py uses the <tt>watch()</tt> method to log default metrics to Weights & Biases. See their [https://docs.wandb.ai full documentation] for more options.
The script wandb-test.py uses the <tt>watch()</tt> method to log default metrics to Weights & Biases. See their [https://docs.wandb.ai full documentation] for more options.


<!--T:14-->
{{File
{{File
   |name=wandb-test.py
   |name=wandb-test.py
Line 62: Line 73:
import torch.backends.cudnn as cudnn
import torch.backends.cudnn as cudnn


<!--T:15-->
import torchvision
import torchvision
import torchvision.transforms as transforms
import torchvision.transforms as transforms
Line 67: Line 79:
from torch.utils.data import DataLoader
from torch.utils.data import DataLoader


<!--T:16-->
import argparse
import argparse


<!--T:17-->
import wandb
import wandb




<!--T:18-->
parser = argparse.ArgumentParser(description='cifar10 classification models, wandb test')
parser = argparse.ArgumentParser(description='cifar10 classification models, wandb test')
parser.add_argument('--lr', default=0.1, help='')
parser.add_argument('--lr', default=0.1, help='')
Line 78: Line 93:
parser.add_argument('--num_workers', type=int, default=0, help='')
parser.add_argument('--num_workers', type=int, default=0, help='')


<!--T:19-->
def main():
def main():
      
      
     args = parser.parse_args()
     args = parser.parse_args()


     print("Starting Wandb...")
     <!--T:20-->
print("Starting Wandb...")


     wandb.init(project="wandb-pytorch-test", config=args)
     <!--T:21-->
wandb.init(project="wandb-pytorch-test", config=args)


     class Net(nn.Module):
     <!--T:22-->
class Net(nn.Module):


       def __init__(self):
       <!--T:23-->
def __init__(self):
           super(Net, self).__init__()
           super(Net, self).__init__()


           self.conv1 = nn.Conv2d(3, 6, 5)
           <!--T:24-->
self.conv1 = nn.Conv2d(3, 6, 5)
           self.pool = nn.MaxPool2d(2, 2)
           self.pool = nn.MaxPool2d(2, 2)
           self.conv2 = nn.Conv2d(6, 16, 5)
           self.conv2 = nn.Conv2d(6, 16, 5)
Line 98: Line 119:
           self.fc3 = nn.Linear(84, 10)
           self.fc3 = nn.Linear(84, 10)


       def forward(self, x):
       <!--T:25-->
def forward(self, x):
           x = self.pool(F.relu(self.conv1(x)))
           x = self.pool(F.relu(self.conv1(x)))
           x = self.pool(F.relu(self.conv2(x)))
           x = self.pool(F.relu(self.conv2(x)))
Line 107: Line 129:
           return x
           return x


     net = Net()
     <!--T:26-->
net = Net()


     transform_train = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
     <!--T:27-->
transform_train = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])


     dataset_train = CIFAR10(root='./data', train=True, download=False, transform=transform_train)
     <!--T:28-->
dataset_train = CIFAR10(root='./data', train=True, download=False, transform=transform_train)


     train_loader = DataLoader(dataset_train, batch_size=args.batch_size, num_workers=args.num_workers)
     <!--T:29-->
train_loader = DataLoader(dataset_train, batch_size=args.batch_size, num_workers=args.num_workers)


     criterion = nn.CrossEntropyLoss()
     <!--T:30-->
criterion = nn.CrossEntropyLoss()
     optimizer = optim.SGD(net.parameters(), lr=args.lr)
     optimizer = optim.SGD(net.parameters(), lr=args.lr)


     wandb.watch(net)
     <!--T:31-->
wandb.watch(net)


     for epoch in range(args.max_epochs):
     <!--T:32-->
for epoch in range(args.max_epochs):


         train(epoch, net, criterion, optimizer, train_loader)
         <!--T:33-->
train(epoch, net, criterion, optimizer, train_loader)




<!--T:34-->
def train(epoch, net, criterion, optimizer, train_loader):
def train(epoch, net, criterion, optimizer, train_loader):


     for batch_idx, (inputs, targets) in enumerate(train_loader):
     <!--T:35-->
for batch_idx, (inputs, targets) in enumerate(train_loader):


       outputs = net(inputs)
       <!--T:36-->
outputs = net(inputs)
       loss = criterion(outputs, targets)
       loss = criterion(outputs, targets)


       optimizer.zero_grad()
       <!--T:37-->
optimizer.zero_grad()
       loss.backward()
       loss.backward()
       optimizer.step()
       optimizer.step()




<!--T:38-->
if __name__=='__main__':
if __name__=='__main__':
   main()
   main()


<!--T:39-->
}}
}}
</translate>
</translate>
Diane27
rsnt_translations
56,420

edits

Retrieved from "https://docs.alliancecan.ca/wiki/Special:MobileDiff/95671"

Navigation menu