Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Performing gradient accumulation with πŸ€— Accelerate

Gradient accumulation is a technique where you can train on bigger batch sizes than your machine would normally be able to fit into memory. This is done by accumulating gradients over several batches, and only stepping the optimizer after a certain number of batches have been performed.

While technically standard gradient accumulation code would work fine in a distributed setup, it is not the most efficient method for doing so and you may experience considerable slowdowns!

In this tutorial you will see how to quickly setup gradient accumulation and perform it with the utilities provided in πŸ€— Accelerate, which can total to adding just one new line of code!

This example will use a very simplistic PyTorch training loop that performs gradient accumulation every two batches:

device = "cuda"
model.to(device)

gradient_accumulation_steps = 2

for index, batch in enumerate(training_dataloader):
    inputs, targets = batch
    inputs = inputs.to(device)
    targets = targets.to(device)
    outputs = model(inputs)
    loss = loss_function(outputs, targets)
    loss = loss / gradient_accumulation_steps
    loss.backward()
    if (index + 1) % gradient_accumulation_steps == 0:
        optimizer.step()
        scheduler.step()
        optimizer.zero_grad()

Converting it to πŸ€— Accelerate

First the code shown earlier will be converted to utilize πŸ€— Accelerate without the special gradient accumulation helper:

+ from accelerate import Accelerator
+ accelerator = Accelerator()

+ model, optimizer, training_dataloader, scheduler = accelerator.prepare(
+     model, optimizer, training_dataloader, scheduler
+ )

  for index, batch in enumerate(training_dataloader):
      inputs, targets = batch
-     inputs = inputs.to(device)
-     targets = targets.to(device)
      outputs = model(inputs)
      loss = loss_function(outputs, targets)
      loss = loss / gradient_accumulation_steps
+     accelerator.backward(loss)
      if (index+1) % gradient_accumulation_steps == 0:
          optimizer.step()
          scheduler.step()
          optimizer.zero_grad()

In its current state, this code is not going to perform gradient accumulation efficiently due to a process called gradient synchronization. Read more about that in the Concepts tutorial!

Letting πŸ€— Accelerate handle gradient accumulation

All that is left now is to let πŸ€— Accelerate handle the gradient accumulation for us. To do so you should pass in a gradient_accumulation_steps parameter to Accelerator, dictating the number of steps to perform before each call to step() and how to automatically adjust the loss during the call to backward():

  from accelerate import Accelerator
- accelerator = Accelerator()
+ accelerator = Accelerator(gradient_accumulation_steps=2)

From here you can use the accumulate() context manager from inside your training loop to automatically perform the gradient accumulation for you! You just wrap it around the entire training part of our code:

- for index, batch in enumerate(training_dataloader):
+ for batch in training_dataloader:
+     with accelerator.accumulate(model):
          inputs, targets = batch
          outputs = model(inputs)

You can remove all the special checks for the step number and the loss adjustment:

- loss = loss / gradient_accumulation_steps
  accelerator.backward(loss)
- if (index+1) % gradient_accumulation_steps == 0:
  optimizer.step()
  scheduler.step()
  optimizer.zero_grad()

As you can see the Accelerator is able to keep track of the batch number you are on and it will automatically know whether to step through the prepared optimizer and how to adjust the loss.

The finished code

Below is the finished implementation for performing gradient accumulation with πŸ€— Accelerate

for batch in training_dataloader:
    with accelerator.accumulate(model):
        inputs, targets = batch
        outputs = model(inputs)
        loss = loss_function(outputs, targets)
        accelerator.backward(loss)
        optimizer.step()
        scheduler.step()
        optimizer.zero_grad()

To learn more about what magic this wraps around, read the Gradient Synchronization concept guide