Multi-GPU training basics in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the main benefit of using multiple GPUs for training a machine learning model?

Using multiple GPUs speeds up training by splitting the work across devices, allowing the model to learn faster from more data at the same time.

beginner

What is 'data parallelism' in multi-GPU training?

Data parallelism means copying the model on each GPU and splitting the input data among them. Each GPU processes its part, then results are combined to update the model.

intermediate

In TensorFlow, which API helps to easily distribute training across multiple GPUs?

TensorFlow's tf.distribute.Strategy API, especially tf.distribute.MirroredStrategy, helps run training on multiple GPUs by managing data splitting and syncing automatically.

intermediate

Why do we need to synchronize gradients across GPUs during multi-GPU training?

Synchronizing gradients ensures all GPUs update the model weights consistently, so the model learns correctly as if training on one device.

advanced

What is a common challenge when using multiple GPUs for training?

A common challenge is communication overhead between GPUs, which can slow down training if not managed well, especially when syncing gradients.

What does tf.distribute.MirroredStrategy do in TensorFlow?

A Copies the model on each GPU and synchronizes updates

B Runs the model only on CPU

C Splits the model layers across GPUs

D Converts the model to a smaller size

MirroredStrategy copies the entire model on each GPU and keeps the weights in sync by mirroring updates.

Which of the following best describes data parallelism?

A Splitting the model across GPUs

B Splitting the input data across GPUs with full model copies

C Using only one GPU for training

D Training different models on each GPU

Data parallelism means each GPU has a full copy of the model and processes a portion of the input data.

Why is gradient synchronization important in multi-GPU training?

A To ensure all GPUs update the model weights the same way

B To increase the batch size

C To reduce the number of GPUs needed

D To speed up data loading

Synchronizing gradients makes sure all GPUs update the model consistently, preventing training errors.

What is a potential downside of multi-GPU training?

A Cannot use batch normalization

B Model accuracy always decreases

C Slower training due to communication overhead

D Only works with small datasets

Communication between GPUs can slow training if not optimized, especially when syncing gradients.

Which TensorFlow API is designed to simplify multi-GPU training?

A tf.keras.Sequential

B tf.function

C tf.data.Dataset

D tf.distribute.Strategy

tf.distribute.Strategy helps distribute training across multiple devices like GPUs.

Explain how multi-GPU training speeds up model training and what role data parallelism plays.

Think about how work is shared across GPUs.

Describe the purpose of gradient synchronization in multi-GPU training and what might happen if it is not done.

Consider how model weights stay the same across GPUs.

Practice

1. What is the main purpose of using tf.distribute.MirroredStrategy() in TensorFlow?

easy

A. To train a model on multiple GPUs simultaneously

B. To reduce the size of the dataset

C. To convert a model to run on CPU only

D. To save the model in a special format

Solution

1. Step 1: Understand the role of MirroredStrategy

   tf.distribute.MirroredStrategy() is designed to distribute training across multiple GPUs to speed up the process.

2. Step 2: Compare options with the purpose

   Options A, B, and C do not relate to multi-GPU training. Only training a model on multiple GPUs simultaneously correctly states the purpose.

3. Final Answer:

   To train a model on multiple GPUs simultaneously -> Option A

4. Quick Check:

   Multi-GPU training = MirroredStrategy usage [OK]

Hint: MirroredStrategy means training on many GPUs at once [OK]

Common Mistakes:

• Thinking it reduces dataset size
• Confusing with CPU-only training
• Assuming it changes model saving format

2. Which of the following is the correct way to start using multi-GPU training in TensorFlow?

easy

A. model = build_model()
strategy = tf.distribute.MirroredStrategy()
model.compile()

B. strategy = tf.distribute.MirroredStrategy()
with strategy.scope():
model = build_model()

C. strategy = tf.distribute.MirroredStrategy()
model = build_model()
model.fit()

D. model = build_model()
model.fit()
strategy = tf.distribute.MirroredStrategy()

Solution

1. Step 1: Identify correct usage of strategy.scope()

   Model building and compiling must happen inside strategy.scope() to enable multi-GPU support.

2. Step 2: Check each option for correct order

   Only strategy = tf.distribute.MirroredStrategy() with strategy.scope(): model = build_model() builds the model inside the strategy scope. Others build or compile outside, which disables multi-GPU benefits.

3. Final Answer:

   strategy = tf.distribute.MirroredStrategy() with strategy.scope(): model = build_model() -> Option B

4. Quick Check:

   Build model inside strategy.scope() = strategy = tf.distribute.MirroredStrategy() with strategy.scope(): model = build_model() [OK]

Hint: Always build and compile inside strategy.scope() [OK]

Common Mistakes:

• Building model before strategy.scope()
• Compiling model outside strategy.scope()
• Calling fit before defining strategy

3. Given the code below, what will be the output shape of the predictions if the input batch size is 64 and there are 2 GPUs?

strategy = tf.distribute.MirroredStrategy()
with strategy.scope():
    model = tf.keras.Sequential([
        tf.keras.layers.Dense(10, activation='softmax')
    ])
    model.compile(optimizer='adam', loss='sparse_categorical_crossentropy')

import numpy as np
inputs = np.random.random((64, 20))
predictions = model.predict(inputs)
print(predictions.shape)

medium

A. (10, 64)

B. (32, 10)

C. (128, 10)

D. (64, 10)

Solution

1. Step 1: Understand batch size and MirroredStrategy behavior

   MirroredStrategy splits the batch across GPUs internally but the output predictions shape matches the input batch size.

2. Step 2: Check model output shape

   The model outputs 10 classes per input. Input batch size is 64, so output shape is (64, 10).

3. Final Answer:

   (64, 10) -> Option D

4. Quick Check:

   Output shape = (batch_size, classes) = (64, 10) [OK]

Hint: Output shape matches input batch size, not multiplied by GPUs [OK]

Common Mistakes:

• Multiplying batch size by number of GPUs
• Confusing output shape dimensions
• Assuming output shape is (10, 64)

4. You wrote this code but get an error: RuntimeError: tf.distribute.MirroredStrategy() must be used with model creation inside its scope. What is the likely mistake?

strategy = tf.distribute.MirroredStrategy()
model = build_model()
with strategy.scope():
    model.compile(optimizer='adam', loss='mse')

medium

A. Model was compiled outside the strategy.scope()

B. MirroredStrategy cannot be used with compile()

C. Model was built outside the strategy.scope()

D. The optimizer 'adam' is not supported

Solution

1. Step 1: Analyze error message and code

   The error says model creation must be inside strategy.scope(). Here, model is built before entering the scope.

2. Step 2: Identify correct fix

   Building the model inside the scope fixes the error. Compiling inside scope is correct.

3. Final Answer:

   Model was built outside the strategy.scope() -> Option C

4. Quick Check:

   Build model inside scope to avoid RuntimeError [OK]

Hint: Build model inside strategy.scope() to fix this error [OK]

Common Mistakes:

• Building model before strategy.scope()
• Thinking compile() causes error
• Blaming optimizer choice

5. You want to train a large model on 4 GPUs using TensorFlow. Which of the following steps correctly applies multi-GPU training and ensures the model can handle larger batch sizes efficiently?

hard

A. Use tf.distribute.MirroredStrategy(), build and compile the model inside its scope, then increase batch size in model.fit()

B. Build the model first, then use tf.distribute.MirroredStrategy() only during model.fit() with default batch size

C. Use tf.distribute.MirroredStrategy() but keep batch size small to avoid memory issues

D. Train the model separately on each GPU without using MirroredStrategy and average weights manually

Solution

1. Step 1: Apply MirroredStrategy correctly

   Use tf.distribute.MirroredStrategy() and build/compile the model inside its scope to enable multi-GPU training.

2. Step 2: Adjust batch size for efficiency

   Increasing batch size in model.fit() leverages multiple GPUs to handle larger data efficiently.

3. Final Answer:

   Use tf.distribute.MirroredStrategy(), build and compile the model inside its scope, then increase batch size in model.fit() -> Option A

4. Quick Check:

   Build inside scope + increase batch size = efficient multi-GPU training [OK]

Hint: Build inside scope and increase batch size for multi-GPU speed [OK]

Common Mistakes:

• Building model outside strategy scope
• Not increasing batch size to use GPUs fully
• Manually averaging weights instead of using MirroredStrategy