I am using clearml to explore learning rates in PyTorch. Tuning task is now DRAFT and data is not showing up on loss monitoring screen.

step1:(This has been a success.)

import torch
import torchvision
import torchvision.transforms as transforms
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from clearml import Task, Logger

task = Task.init(project_name="CNN Experiments", task_name="non tune 3 class cnn")
logger = Logger.current_logger()

transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform)
testset = torchvision.datasets.CIFAR10(root='./data', train=False, download=True, transform=transform)

#extract 3 class
train_indices = [i for i, (x, y) in enumerate(trainset) if y < 3]
test_indices = [i for i, (x, y) in enumerate(testset) if y < 3]

trainloader = torch.utils.data.DataLoader(
    torch.utils.data.Subset(trainset, train_indices),
    batch_size=64, shuffle=True, num_workers=2
)

testloader = torch.utils.data.DataLoader(
    torch.utils.data.Subset(testset, test_indices),
    batch_size=64, shuffle=False, num_workers=2
)

# tune param
configuration_dict = {"base_lr": 0.001,'dropout': 0.25,'number_of_epochs': 3}
configuration_dict = task.connect(configuration_dict) 

class SimpleCNN(nn.Module):
    def __init__(self):
        #super(SimpleCNN, self).__init__()
        super().__init__()
        self.conv1 = nn.Conv2d(3, 16, 3, padding=1)  # 3x32x32 -> 16x32x32
        self.conv2 = nn.Conv2d(16, 32, 3, padding=1)  # 16x32x32 -> 32x32x32
        self.conv3 = nn.Conv2d(32, 64, 3, padding=1)  # 32x32x32 -> 64x32x32
        self.pool = nn.MaxPool2d(2, 2) 
        self.fc1 = nn.Linear(64 * 4 * 4, 64)
        self.dorpout = nn.Dropout(p=configuration_dict.get('dropout', 0.25))
        self.fc2 = nn.Linear(64, 3)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))  
        x = self.pool(F.relu(self.conv2(x)))  
        x = self.pool(F.relu(self.conv3(x)))  
        x = torch.flatten(x, 1)


        x = F.relu(self.fc1(x))
        x = self.fc2(x)
        return x

model = SimpleCNN()
criterion = nn.CrossEntropyLoss() 
optimizer = optim.SGD(model.parameters(), lr=configuration_dict.get('base_lr', 0.001), momentum=0.9)  


for epoch in range(configuration_dict.get('number_of_epochs', 10)): 
    running_loss = 0.0
    for i, data in enumerate(trainloader, 0):
        inputs, labels = data
        optimizer.zero_grad()
        outputs = model(inputs)  
        loss = criterion(outputs, labels) 
        loss.backward() 
        optimizer.step() 

        running_loss += loss.item()
        if i % 100 == 99:  
            print(f"[{epoch + 1}, {i + 1}] loss: {running_loss / 100:.3f}")
            running_loss = 0.0

        #############
    logger.report_scalar(title="train", series="loss", iteration=epoch, value=loss.item())
    print("send loss")

print("Finished Training")

PATH = './cifar_net.pth'
torch.save(model.state_dict(), PATH)



correct = 0
total = 0
with torch.no_grad(): 
    for data in testloader:
        images, labels = data
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()

print(f"Accuracy on test data: {100 * correct / total:.2f}%")

In the above code, loss is displayed in the scalars panel.
However, when the following tuning code is executed, the loss for the parameter combination is not recorded.

Is there a problem?

step2:(tune)

from clearml import Task, Logger
from clearml.automation import UniformParameterRange, UniformIntegerParameterRange
from clearml.automation import DiscreteParameterRange
from clearml.automation import GridSearch, RandomSearch, HyperParameterOptimizer
from clearml.automation.optuna import OptimizerOptuna

def job_complete_callback(
    job_id,                 # type: str
    objective_value,        # type: float
    objective_iteration,    # type: int
    job_parameters,         # type: dict
    top_performance_job_id  # type: str
):
    print('Job completed!', job_id, objective_value, objective_iteration, job_parameters)
    if job_id == top_performance_job_id:
        print('WOOT WOOT we broke the record! Objective reached {}'.format(objective_value))

task = Task.init(project_name="CNN tuner", 
    task_name="tune",
    task_type=Task.TaskTypes.optimizer,
    #reuse_last_task_id=False
    )

#base_task_id is step1-task-id
an_optimizer = HyperParameterOptimizer(
    base_task_id='9e9ce547---------2d167936',
    hyper_parameters=[
        UniformIntegerParameterRange('number_of_epochs', min_value=2, max_value=12, step_size=2),
        UniformParameterRange('dropout', min_value=0, max_value=0.5, step_size=0.05),
        UniformParameterRange('base_lr', min_value=0.00025, max_value=0.01, step_size=0.00025),
    ],
    objective_metric_title='train',
    objective_metric_series='loss',
    objective_metric_sign='min',

    optimizer_class=OptimizerOptuna,
    # Configuring optimization parameters
    #execution_queue='dan_queue',  
    # queue to schedule the experiments for execution
    max_number_of_concurrent_tasks=20,  # number of concurrent experiments
    optimization_time_limit=6.,  # set the time limit for the optimization process
    compute_time_limit=6,  # set the compute time limit (sum of execution time on all machines)
    total_max_jobs=20,  # set the maximum number of experiments for the optimization. 
                        # Converted to total number of iteration for OptimizerBOHB
    min_iteration_per_job=150,  # minimum number of iterations per experiment, till early stopping
    max_iteration_per_job=1500,  # maximum number of iterations per experiment

    )

an_optimizer.set_report_period(0.2) 
an_optimizer.start()
an_optimizer.wait()
an_optimizer.stop()

top_exp = an_optimizer.get_top_experiments(top_k=3)
print([t.id for t in top_exp])

print("all fin")

The ideal goal would be to obtain a figure comparing the performance of the parameters, such as the following URL

https://clear.ml/docs/latest/assets/images/webapp_compare_11-ad1348542d5b42a0cce427e91e9ec2e5.png

enter image description here

I have looked at the official youtube and run the github exsamples but could not achieve it with my code.

https://github.com/allegroai/clearml/tree/master/examples