wandb but how to sweeps?

.gitignore

@@ -1,4 +1,6 @@
 **/data/
 **/__pycache__/
 .venv/
+**/logs/
+**/wandb/
 

challenges-in-representation-learning-facial-expression-recognition-challenge/main.py

@@ -16,20 +16,13 @@
 from torch.nn import functional as F
 from tqdm import tqdm
 
+import wandb
+
 from torch.utils.tensorboard import SummaryWriter
 
 import csv, pickle
 from pathlib import Path
 
-def grouper(n, iterable):
-    bad = []
-    for i in iterable:
-        bad.append(i)
-        if len(bad) == n:
-            yield torch.stack(bad)
-            bad = []
-    yield torch.stack(bad)
-
 def plot_grad_flow_bars(named_parameters):
     # from @jemoka inscriptio gc
     ave_grads = []
@@ -57,6 +50,15 @@ def plot_grad_flow_bars(named_parameters):
     plt.show()
 
 def load_data():
+    def grouper(n, iterable):
+        bad = []
+        for i in iterable:
+            bad.append(i)
+            if len(bad) == n:
+                yield torch.stack(bad)
+                bad = []
+        yield torch.stack(bad)
+
     if Path(DATAPATH+'.pkl').exists():
         print('found cached data; loading it...')
         with open(DATAPATH+'.pkl', 'rb') as rf:
@@ -91,11 +93,11 @@ def load_data():
 class Net(nn.Module):
     def __init__(self):
         super().__init__()
-        self.conv1 = nn.Conv2d(1, 6, 5)     # -> 6  x 44x44
-        self.pool = nn.MaxPool2d(2, 2)      # -> 6  x 22x22;    default stride = kernel_size
-        self.conv2 = nn.Conv2d(6, 16, 5)    # -> 16 x 18x18
-        # pool again                        # -> 16 x 9 x 9
-        self.full1 = nn.Linear(16 * 9*9, 200)
+        self.conv1 = nn.Conv2d(1, 10, 5)    # -> 10 x 44x44
+        self.pool = nn.MaxPool2d(2, 2)      # -> 10 x 22x22;    default stride = kernel_size
+        self.conv2 = nn.Conv2d(10, 20, 5)   # -> 20 x 18x18
+        # pool again                        # -> 20 x 9 x 9
+        self.full1 = nn.Linear(20 * 9*9, 200)
         self.full2 = nn.Linear(200, 70)
         self.full3 = nn.Linear(70, 7)
         self.final = nn.Softmax(dim=1)
@@ -110,25 +112,31 @@ def forward(self, x):
         return self.final(x)
 
 if __name__ == '__main__':
+    wandb.init(project='facial expression recognition')
     print(f'pytorch version is {torch.__version__}')
 
+    device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+    print(f'training on device {device}')
+
     data = list(load_data())
 
     net = Net()
     print(net)
     print(f'parameter count: {len(list(net.parameters()))}')
+    net.to(device)
 
     criterion = nn.CrossEntropyLoss()
     optimizer = optim.Adam(net.parameters(), lr=LEARNING_RATE)
 
+
     if SHOULD_LOG:
-        writer = SummaryWriter(LOGS_DIR) # TODO: with statement
+        wandb.watch(net)
+        # writer = SummaryWriter(LOGS_DIR) # TODO: with statement
 
     with tqdm(total=EPOCHS*len(data)) as pbar:
         for epoch in range(EPOCHS):
-            running_loss = 0.               # TODO: nanny
             for i, samp in enumerate(data):
-                img, cls = samp
+                img, cls = samp[0].to(device), samp[1].to(device)
                 # onehot = F.one_hot(cls, len(DATA_CLASSES)).type(torch.float32)
 
                 optimizer.zero_grad()
@@ -140,12 +148,14 @@ def forward(self, x):
                 #     plot_grad_flow_bars(net.named_parameters())
                 optimizer.step()
 
-                running_loss += loss.item()
                 pbar.update(1)
-                pbar.set_description(f'step {epoch*len(data)+i}; loss {loss.item():.3f}')
-                if SHOULD_LOG:
-                    writer.add_scalar('loss', loss.item(), epoch*len(data)+i)
+                if (epoch*len(data)+i) % int(1e2) == 0:
+                    pbar.set_description(f'step {epoch*len(data)+i}; loss {loss.item():.3f}')
+                    if SHOULD_LOG:
+                        wandb.log({'loss': loss})
+                        # writer.add_scalar('loss', loss.item(), epoch*len(data)+i)
 
     if SHOULD_LOG:
-        writer.close()
+        pass
+        # writer.close()
 

requirements.txt

@@ -0,0 +1,41 @@
+absl-py==0.13.0
+cachetools==4.2.2
+certifi==2021.5.30
+chardet==4.0.0
+cycler==0.10.0
+google-auth==1.32.0
+google-auth-oauthlib==0.4.4
+grpcio==1.38.0
+idna==2.10
+kiwisolver==1.3.1
+llvmlite==0.36.0
+Markdown==3.3.4
+matplotlib==3.4.2
+numba==0.53.1
+numpy==1.20.3
+oauthlib==3.1.1
+pandas==1.2.4
+Pillow==8.2.0
+protobuf==3.17.3
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pyparsing==2.4.7
+PyQt5==5.15.4
+PyQt5-Qt5==5.15.2
+PyQt5-sip==12.9.0
+python-dateutil==2.8.1
+pytz==2021.1
+requests==2.25.1
+requests-oauthlib==1.3.0
+rsa==4.7.2
+six==1.16.0
+tensorboard==2.5.0
+tensorboard-data-server==0.6.1
+tensorboard-plugin-wit==1.8.0
+torch==1.9.0+cu111
+torchaudio==0.9.0
+torchvision==0.10.0+cu111
+tqdm==4.61.1
+typing-extensions==3.10.0.0
+urllib3==1.26.5
+Werkzeug==2.0.1