Pretrained Transformer Image Classification Task¶

In [ ]:

# openml imports
import openml
import openml_pytorch as op
from openml_pytorch.callbacks import TestCallback
from openml_pytorch.metrics import accuracy
from openml_pytorch.trainer import convert_to_rgb

# pytorch imports
from torch.utils.tensorboard.writer import SummaryWriter
from torchvision.transforms import Compose, Resize, ToPILImage, ToTensor, Lambda
import torchvision
import torch

# other imports
import logging
import warnings

# set up logging
openml.config.logger.setLevel(logging.DEBUG)
openml_pytorch.config.logger.setLevel(logging.DEBUG)
warnings.simplefilter(action='ignore')

# openml imports import openml import openml_pytorch as op from openml_pytorch.callbacks import TestCallback from openml_pytorch.metrics import accuracy from openml_pytorch.trainer import convert_to_rgb # pytorch imports from torch.utils.tensorboard.writer import SummaryWriter from torchvision.transforms import Compose, Resize, ToPILImage, ToTensor, Lambda import torchvision import torch # other imports import logging import warnings # set up logging openml.config.logger.setLevel(logging.DEBUG) openml_pytorch.config.logger.setLevel(logging.DEBUG) warnings.simplefilter(action='ignore')

Data¶

Define image transformations¶

In [ ]:

transform = Compose(
    [
        ToPILImage(),  # Convert tensor to PIL Image to ensure PIL Image operations can be applied.
        Lambda(convert_to_rgb),  # Convert PIL Image to RGB if it's not already.
        Resize((64, 64)),  # Resize the image.
        ToTensor(),  # Convert the PIL Image back to a tensor.
    ]
)

transform = Compose( [ ToPILImage(), # Convert tensor to PIL Image to ensure PIL Image operations can be applied. Lambda(convert_to_rgb), # Convert PIL Image to RGB if it's not already. Resize((64, 64)), # Resize the image. ToTensor(), # Convert the PIL Image back to a tensor. ] )

Configure the Data Module and Choose a Task¶

• Make sure the data is present in the file_dir directory, and the filename_col is correctly set along with this column correctly pointing to where your data is stored.

In [ ]:

data_module = op.OpenMLDataModule(
    type_of_data="image",
    file_dir="datasets",
    filename_col="image_path",
    target_mode="categorical",
    target_column="label",
    batch_size=64,
    transform=transform,
)

# Download the OpenML task for tiniest imagenet
task = openml.tasks.get_task(363295)

data_module = op.OpenMLDataModule( type_of_data="image", file_dir="datasets", filename_col="image_path", target_mode="categorical", target_column="label", batch_size=64, transform=transform, ) # Download the OpenML task for tiniest imagenet task = openml.tasks.get_task(363295)

Model¶

In [ ]:

# Example model. You can do better :)
import torchvision.models as models
import torch.nn as nn

# Load the pre-trained model
model = models.efficientnet_b0(pretrained=True)

# Modify the last fully connected layer to the required number of classes
num_classes = 200
in_features = model.classifier[-1].in_features
model.classifier = nn.Sequential(
    nn.Dropout(p=0.2, inplace=True),
    nn.Linear(in_features, num_classes),
)

# Optional: If you're fine-tuning, you may want to freeze the pre-trained layers
# for param in model.parameters():
#     param.requires_grad = False

# # If you want to train the last layer only (the newly added layer)
# for param in model.fc.parameters():
#     param.requires_grad = True

# Example model. You can do better :) import torchvision.models as models import torch.nn as nn # Load the pre-trained model model = models.efficientnet_b0(pretrained=True) # Modify the last fully connected layer to the required number of classes num_classes = 200 in_features = model.classifier[-1].in_features model.classifier = nn.Sequential( nn.Dropout(p=0.2, inplace=True), nn.Linear(in_features, num_classes), ) # Optional: If you're fine-tuning, you may want to freeze the pre-trained layers # for param in model.parameters(): # param.requires_grad = False # # If you want to train the last layer only (the newly added layer) # for param in model.fc.parameters(): # param.requires_grad = True

Train your model on the data¶

• Note that by default, OpenML runs a 10 fold cross validation on the data. You cannot change this for now.

In [ ]:

trainer = op.OpenMLTrainerModule(
    experiment_name= "Tiny ImageNet",
    data_module=data_module,
    verbose=True,
    epoch_count=1,
    metrics= [accuracy],
    # remove the TestCallback when you are done testing your pipeline. Having it here will make the pipeline run for a very short time.
    callbacks=[
        TestCallback,
    ],
    opt=torch.optim.Adam,
)
op.config.trainer = trainer
run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)

trainer = op.OpenMLTrainerModule( experiment_name= "Tiny ImageNet", data_module=data_module, verbose=True, epoch_count=1, metrics= [accuracy], # remove the TestCallback when you are done testing your pipeline. Having it here will make the pipeline run for a very short time. callbacks=[ TestCallback, ], opt=torch.optim.Adam, ) op.config.trainer = trainer run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)

View information about your run¶

Learning rate and loss plot¶

In [ ]:

trainer.plot_loss()

trainer.plot_loss()

In [ ]:

trainer.plot_lr()

trainer.plot_lr()

Class labels¶

In [ ]:

trainer.model_classes

trainer.model_classes

Model Vizualization¶

• Sometimes you may want to visualize the model. You can either use netron or tensorboard for this purpose.

Netron¶

In [ ]:

trainer.export_to_netron()

trainer.export_to_netron()

Tensorboard¶

• By default, openml will log the tensorboard logs in the tensorboard_logs directory. You can view the logs by running tensorboard --logdir tensorboard_logs in the terminal.

Publish your model to OpenML¶

• This is Optional, but publishing your model to OpenML will allow you to track your experiments and compare them with others.
• Make sure to set your apikey first.
  • You can find your apikey on your OpenML account page.

In [ ]:

openml.config.apikey = ''
run = op.add_experiment_info_to_run(run=run, trainer=trainer) 
run.publish()

openml.config.apikey = '' run = op.add_experiment_info_to_run(run=run, trainer=trainer) run.publish()