slides.md

title	tags	description	slideOptions
Experiment management with lightning and tensorboard	Talk, pytorch lightning, tensorboard	Oceanix workshop.	spotlight enabled true

Workshop

Pytorch lightning and tensorboard for better experiment management

---

Why

Demonstrate how the right tools can make your life easier https://media.giphy.com/media/0qaV5zc6KBMxrnELxK/giphy.gif

---

Why Lightning ?

Engineering vs Research : Spending time on what matters

• Infra (cpu, gpu, cluster)
• managing logs
• organizing code (software engineering)
• saving/loading models

---

Why tensorboard ?

Tracking and comparing results of different experiments

A demo is worth a thousand words

---

Plan

• Lightning code walkthrough
• Workshop exercises description
• To your keyboards

---

Lightning code walkthrough

class MNISTModel(LightningModule):

    def __init__(self):
        super().__init__()
        self.l1 = torch.nn.Linear(28 * 28, 10)

    def forward(self, x):
        return torch.relu(self.l1(x.view(x.size(0), -1)))

    def training_step(self, batch, batch_nb):
        x, y = batch
        loss = F.cross_entropy(self(x), y)
        return loss

    def configure_optimizers(self):
        return torch.optim.Adam(self.parameters(), lr=0.02)



# Init our model
mnist_model = MNISTModel()

# Init DataLoader from MNIST Dataset
train_ds = MNIST(PATH_DATASETS, train=True, download=True, transform=transforms.ToTensor())
train_loader = DataLoader(train_ds, batch_size=BATCH_SIZE)

# Initialize a trainer
trainer = Trainer(
    max_epochs=3,
)

# Train the model 
trainer.fit(mnist_model, train_loader)

---

Let's start

The repo for the workshop:

https://github.com/quentinf00/lightning-tensorboard-workshop