GPU tensors (to, cuda) in PyTorch - Model Metrics & Evaluation

Metrics & Evaluation - GPU tensors (to, cuda)

Which metric matters for GPU tensors (to, cuda) and WHY

When using GPU tensors in PyTorch, the main goal is to speed up model training and inference. The key metric to watch is training time or inference time. Faster times mean the GPU is helping. Accuracy or loss stay the same whether on CPU or GPU, but speed improves. So, measuring how much faster your model runs on GPU compared to CPU is the most important metric here.

Confusion matrix or equivalent visualization

GPU tensors do not affect prediction correctness directly, so confusion matrix is not relevant here. Instead, consider a simple timing comparison:

CPU time: 10 seconds
GPU time: 2 seconds
Speedup: 5x faster
    

This shows the benefit of moving tensors to GPU using to('cuda') or cuda().

Precision vs Recall tradeoff (or equivalent) with concrete examples

For GPU tensors, the tradeoff is between speed and resource use. Using GPU speeds up training but uses more power and GPU memory. Using CPU is slower but uses less power and memory.

Example: Training a model on CPU takes 10 minutes, on GPU takes 2 minutes but uses more electricity and GPU memory. Choose GPU if speed is critical, CPU if resources are limited.

What "good" vs "bad" metric values look like for this use case

Good: GPU training time is significantly less than CPU time (e.g., 5x faster). Model accuracy and loss remain consistent.

Bad: No speed improvement or slower training on GPU due to overhead or incorrect tensor placement. Model errors due to mixing CPU and GPU tensors.

Metrics pitfalls

• Not moving all tensors and model parts to the same device causes errors.
• Measuring accuracy or loss differences caused by device change is misleading; device does not affect correctness.
• Ignoring GPU memory limits can cause out-of-memory errors, stopping training.
• Overhead of moving data to GPU can make small models slower on GPU.

Self-check

Your model runs with 98% accuracy on CPU in 10 minutes. On GPU, accuracy is still 98% but training takes 12 minutes. Is this good?

Answer: No, the GPU training is slower, which means the GPU is not used efficiently. Check if tensors and model are properly moved to GPU and if GPU memory is sufficient.

Key Result

GPU tensors improve training speed significantly when used correctly, without affecting accuracy.

Practice

1. What does the PyTorch method tensor.to('cuda') do?

easy

A. Moves the tensor to the GPU for faster computation

B. Converts the tensor to a CPU tensor

C. Changes the tensor data type to float

D. Deletes the tensor from memory

Solution

1. Step 1: Understand the purpose of to('cuda')

   This method moves the tensor from CPU memory to GPU memory to speed up calculations.

2. Step 2: Recognize the effect on computation

   By moving the tensor to GPU, PyTorch can use GPU cores for faster processing.

3. Final Answer:

   Moves the tensor to the GPU for faster computation -> Option A

4. Quick Check:

   to('cuda') = move tensor to GPU [OK]

Hint: to('cuda') means move tensor to GPU memory [OK]

Common Mistakes:

• Thinking it converts data type
• Assuming it moves tensor to CPU
• Believing it deletes the tensor

2. Which of the following is the correct syntax to move a tensor named x to GPU in PyTorch?

easy

A. x.to('cpu')

B. x.gpu()

C. x.to('cuda')

D. x.cuda('cpu')

Solution

1. Step 1: Identify correct method to move tensor to GPU

   The method to('cuda') moves tensor to GPU.

2. Step 2: Check each option

   x.to('cuda') uses to('cuda'), which is correct. Others either move to CPU or use invalid methods.

3. Final Answer:

   x.to('cuda') -> Option C

4. Quick Check:

   Correct syntax for GPU move = x.to('cuda') [OK]

Hint: Use .to('cuda') to move tensor to GPU [OK]

Common Mistakes:

• Using .to('cpu') instead of 'cuda'
• Calling .cuda('cpu') which is invalid
• Using non-existent .gpu() method

3. What will be the output device of the tensor after running this code?

import torch
x = torch.tensor([1, 2, 3])
if torch.cuda.is_available():
    x = x.cuda()
print(x.device)

medium

A. cuda:0

B. cuda:1

C. cpu

D. None

Solution

1. Step 1: Check if GPU is available

   The code checks torch.cuda.is_available(). If True, it moves tensor to GPU device 0 by default.

2. Step 2: Understand x.device output

   After x.cuda(), the tensor is on GPU device 0, so x.device prints cuda:0.

3. Final Answer:

   cuda:0 -> Option A

4. Quick Check:

   cuda() moves tensor to cuda:0 device [OK]

Hint: cuda() moves tensor to GPU device 0 by default [OK]

Common Mistakes:

• Assuming device is 'cpu' after cuda()
• Thinking cuda() moves to cuda:1 by default
• Expecting None as device

4. Identify the error in this code snippet that tries to move a tensor to GPU:

import torch
x = torch.tensor([1, 2, 3])
x = x.to('gpu')
print(x.device)

medium

A. Tensor cannot be moved to GPU

B. Tensor must be converted to float before moving

C. Missing .cuda() call after .to()

D. The device string 'gpu' is invalid; should be 'cuda'

Solution

1. Step 1: Check device string in to() method

   The string 'gpu' is not recognized by PyTorch; the correct string is 'cuda'.

2. Step 2: Understand correct usage

   Using 'cuda' moves tensor to GPU. Using 'gpu' causes a runtime error.

3. Final Answer:

   The device string 'gpu' is invalid; should be 'cuda' -> Option D

4. Quick Check:

   Use 'cuda' not 'gpu' in to() [OK]

Hint: Use 'cuda' not 'gpu' as device string [OK]

Common Mistakes:

• Using 'gpu' instead of 'cuda'
• Thinking tensor must be float to move
• Missing .cuda() call after .to()

5. You want to write code that safely moves a tensor x to GPU if available, otherwise keeps it on CPU. Which code snippet correctly does this?

hard

A. x = x.to('cuda') if torch.cuda.is_available() else x.to('gpu')

B. if torch.cuda.is_available():
x = x.to('cuda')
else:
x = x.to('cpu')

C. x = x.cuda() if torch.cuda.is_available() else x.cuda()

D. x = x.to('cpu') if torch.cuda.is_available() else x.to('cuda')

Solution

1. Step 1: Understand device check logic

   We must check if GPU is available using torch.cuda.is_available().

2. Step 2: Apply correct device move

   If GPU is available, move tensor to 'cuda', else keep on 'cpu'. if torch.cuda.is_available(): x = x.to('cuda') else: x = x.to('cpu') does this correctly.

3. Final Answer:

   if torch.cuda.is_available(): x = x.to('cuda') else: x = x.to('cpu') -> Option B

4. Quick Check:

   Check GPU availability before moving tensor [OK]

Hint: Check cuda availability before moving tensor [OK]

Common Mistakes:

• Using invalid device string 'gpu'
• Moving to cuda even if unavailable
• Swapping cpu and cuda in else