This is a collection of distributed training experiments instrumented with Weights & Biases.
- basic data parallelism in Keras (multi_gpu_model)
- Tensorflow Distribtued Strategy
- multi_gpu_keras.py: finetune a small CNN on 6-12K photos from iNaturalist
- finetune_experiments.py: finetune large pretrained models (Xception, ResNet, Inception) on iNaturalist