MLOps Engineer
Deep learning and PyTorch questions
18 questions on Deep learning and PyTorch for mlops engineers, easiest first.
Easy 4
Medium 13
- Why bf16 is easier than fp16medium
- Allocated and reserved memorymedium
- empty_cache does not helpmedium
- Memory grows each training stepmedium
- What kernel fusion savesmedium
- What torch.compile doesmedium
- torch.backends.cudnn.benchmarkmedium
- Memory format for CNNsmedium
- Logging slows the training loopmedium
- Writing a custom fused kernelmedium
- TensorRT engine fails on new GPUsmedium
- First requests after deploy are slowmedium
- Variable input lengths in TensorRTmedium