Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
Benchmark of Popular Deep Learning Models
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
Memory
- ResNet152 inferencing requies about (a little bit less than) 2G GPU memory and training can be done with 8G GPU memory (need to set a small batch size). Deeper ResNet …
LightGBM on GPU
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
https://pypi.org/project/lightgbm/#build-gpu-version
https://github.com/microsoft/LightGBM/blob/master/docs/Installation-Guide.rst#build-gpu-version
https://www.kaggle.com/vinhnguyen/gpu-acceleration-for-lightgbm
Microsoft's Example Dockerfile for GPU version of LightGBM …
Tips on TensorFlow
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
Installation
The article Stop Installing Tensorflow using pip for performance sake! suggest installing TensorFlow using conda instead of pip as the version installed by conda leverages Intel Math Kernel Library and …
Use PyTorch on GPU
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
PyTorch on GPU
https://pytorch.org/docs/master/notes/cuda.html
You can use the command torch.cuda.is_available()
to check whether GPU is available for PyTorch.
Details of GPUs can …
Models in Torchvision and Ways to Finetune Them
inception_v3 requires an input of (299, 299)
while other models requires an input of (224, 224).
Due to adaptive pooling used in some models,
they can run on varying sized intput without throwing errors
(but the results are usually not correct).
You have to resize/crop an image to be the right input size
(and then other necessary transformations, e.g., to_tensor