Gradient accumulation in PyTorch - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is gradient accumulation in training neural networks?

Gradient accumulation is a technique where gradients from multiple mini-batches are added together before updating the model weights. This simulates a larger batch size without needing more memory.

beginner

Why use gradient accumulation instead of increasing batch size directly?

Because hardware memory limits may prevent using large batches. Gradient accumulation lets you use small batches but still get the effect of a large batch by summing gradients over steps.

intermediate

In PyTorch, what must you do before calling loss.backward() when using gradient accumulation?

You should NOT call optimizer.zero_grad() every mini-batch. Instead, zero gradients only after the accumulated steps to keep adding gradients.

intermediate

How do you update model weights when using gradient accumulation with an accumulation step of 4?

You call optimizer.step() and optimizer.zero_grad() only every 4 mini-batches after accumulating gradients from each batch.

intermediate

What is a simple PyTorch code pattern for gradient accumulation?

For each mini-batch: compute loss, call loss.backward() without zeroing grads, then every N batches call optimizer.step() and optimizer.zero_grad() to update weights and reset gradients.

What does gradient accumulation help with?

A Increasing learning rate automatically

B Reducing the number of model parameters

C Simulating larger batch sizes without extra memory

D Stopping training early

Gradient accumulation sums gradients over multiple batches to simulate a larger batch size without needing more memory.

When using gradient accumulation, when should you call optimizer.zero_grad()?

A After accumulating gradients over several mini-batches

B After every mini-batch

C Before every mini-batch

D Never

You zero gradients only after accumulating over the desired number of mini-batches to keep adding gradients.

If your accumulation step is 3, how often do you call optimizer.step()?

A Every mini-batch

B Every 3 mini-batches

C Every 2 mini-batches

D Only once at the end

You update weights every 3 mini-batches after accumulating gradients from each.

Which of these is NOT a benefit of gradient accumulation?

A Improves model accuracy by default

B Allows training with large effective batch sizes on limited memory

C Helps stabilize training by simulating bigger batches

D Reduces memory usage compared to large batch training

Gradient accumulation helps with batch size and memory but does not automatically improve accuracy.

What happens if you forget to call optimizer.zero_grad() after accumulation?

A Training stops immediately

B Nothing, training continues normally

C Model weights reset to initial values

D Gradients keep accumulating, possibly causing incorrect updates

Not zeroing gradients causes them to accumulate indefinitely, leading to wrong updates.

Explain how gradient accumulation works and why it is useful in training neural networks.

Think about how small batches can add up to a big batch effect.

Describe the changes needed in a PyTorch training loop to implement gradient accumulation.

Focus on when to zero gradients and when to update weights.

Practice

1. What is the main purpose of gradient accumulation in PyTorch training?

easy

A. To reduce the number of model parameters

B. To increase the learning rate automatically

C. To change the model architecture dynamically

D. To split a large batch into smaller steps to save memory

Solution

1. Step 1: Understand gradient accumulation concept

   Gradient accumulation splits a large batch into smaller mini-batches to avoid memory overflow.

2. Step 2: Identify the main benefit

   This technique helps save GPU memory by processing smaller batches sequentially before updating weights.

3. Final Answer:

   To split a large batch into smaller steps to save memory -> Option D

4. Quick Check:

   Gradient accumulation = batch splitting for memory saving [OK]

Hint: Gradient accumulation saves memory by splitting batches [OK]

Common Mistakes:

• Thinking it changes learning rate
• Confusing with model size reduction
• Assuming it alters model layers

2. Which PyTorch code snippet correctly implements gradient accumulation over 4 steps?

easy

A. for i, data in enumerate(dataloader):
outputs = model(data)
loss = criterion(outputs, labels)
optimizer.step()
loss.backward()
optimizer.zero_grad()

B. for i, data in enumerate(dataloader):
outputs = model(data)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
if (i + 1) % 4 == 0:
optimizer.zero_grad()

C. for i, data in enumerate(dataloader):
outputs = model(data)
loss = criterion(outputs, labels)
loss.backward()
if (i + 1) % 4 == 0:
optimizer.step()
optimizer.zero_grad()

D. for i, data in enumerate(dataloader):
outputs = model(data)
loss = criterion(outputs, labels)
optimizer.zero_grad()
loss.backward()
if (i + 1) % 4 == 0:
optimizer.step()

Solution

1. Step 1: Identify correct order of operations

   Gradient accumulation requires calling loss.backward() each step, then optimizer.step() and zero_grad() after accumulating gradients.

2. Step 2: Check modulo condition for accumulation

   Only after 4 mini-batches (i+1 % 4 == 0) should optimizer.step() and zero_grad() be called to update weights and reset gradients.

3. Final Answer:

   Code that calls loss.backward() every step and optimizer.step() plus zero_grad() every 4 steps -> Option C

4. Quick Check:

   Backward every step, step and zero_grad every 4 steps [OK]

Hint: Call backward every step, step and zero_grad every N steps [OK]

Common Mistakes:

• Calling optimizer.step() before loss.backward()
• Resetting gradients too early
• Not using modulo condition correctly

3. Given this PyTorch code with gradient accumulation steps=2, what is the printed loss after 4 batches?

optimizer.zero_grad()
for i, (inputs, labels) in enumerate(dataloader):
    outputs = model(inputs)
    loss = criterion(outputs, labels) / 2
    loss.backward()
    if (i + 1) % 2 == 0:
        optimizer.step()
        optimizer.zero_grad()
    print(f"Batch {i+1} loss: {loss.item()}")

medium

A. Loss values cause runtime error due to division

B. Loss values divided by 2 printed for each batch

C. Loss values accumulated and printed only every 2 batches

D. Loss values printed without division

Solution

1. Step 1: Understand loss scaling in code

   Loss is divided by 2 before backward to average gradients over 2 accumulation steps.

2. Step 2: Check print statement inside loop

   Loss.item() prints the scaled loss for each batch, so printed values are divided by 2 every batch.

3. Final Answer:

   Loss values divided by 2 printed for each batch -> Option B

4. Quick Check:

   Loss scaled by 2 and printed each batch [OK]

Hint: Loss is scaled before backward, printed as scaled every batch [OK]

Common Mistakes:

• Thinking loss prints only every 2 batches
• Ignoring loss division by 2
• Assuming runtime error from division

4. Identify the error in this gradient accumulation code snippet:

optimizer.zero_grad()
for i, (inputs, labels) in enumerate(dataloader):
    outputs = model(inputs)
    loss = criterion(outputs, labels)
    loss.backward()
    if i % 4 == 0:
        optimizer.step()
        optimizer.zero_grad()

medium

A. optimizer.step() and zero_grad() are called too early at i=0

B. loss.backward() is missing gradient scaling

C. optimizer.zero_grad() should be called before loss.backward()

D. The loop should use (i + 1) % 4 == 0 instead of i % 4 == 0

Solution

1. Step 1: Analyze the modulo condition

   Condition i % 4 == 0 triggers optimizer.step() and zero_grad() at i=0, which is before any gradients accumulate.

2. Step 2: Understand correct timing for optimizer step

   Optimizer step should happen after 4 batches, so condition should be (i + 1) % 4 == 0 to avoid early update.

3. Final Answer:

   optimizer.step() and zero_grad() are called too early at i=0 -> Option A

4. Quick Check:

   Modulo condition triggers step too early [OK]

Hint: Use (i + 1) % N == 0 to delay step until after N batches [OK]

Common Mistakes:

• Calling optimizer.step() at first batch
• Confusing zero_grad() timing
• Ignoring off-by-one in modulo condition

5. You want to train a large model on a GPU with limited memory using gradient accumulation with batch size 64 split into 8 steps. Which approach correctly adjusts the learning rate and loss scaling to keep training stable?

hard

A. Divide the loss by 8 before backward and keep the original learning rate

B. Multiply the learning rate by 8 and do not scale the loss

C. Divide the learning rate by 8 and multiply the loss by 8

D. Keep loss and learning rate unchanged, just accumulate gradients

Solution

1. Step 1: Understand gradient accumulation effect on loss

   Splitting batch into 8 steps means gradients accumulate over 8 mini-batches, so loss must be divided by 8 to average gradients correctly.

2. Step 2: Adjust learning rate accordingly

   Learning rate stays the same because gradient accumulation simulates a larger batch size without changing optimization dynamics.

3. Final Answer:

   Divide the loss by 8 before backward and keep the original learning rate -> Option A

4. Quick Check:

   Loss scaled down, learning rate unchanged for stable accumulation [OK]

Hint: Scale loss down by steps, keep learning rate same [OK]

Common Mistakes:

• Scaling learning rate instead of loss
• Not scaling loss causing gradient explosion
• Assuming no changes needed for accumulation