examples

Catalyst examples

DL notebooks

Tutorials

1. tutorial – classification
   • dataset preparation (raw images -> train/valid/infer splits)
   • augmentations usage example
   • pretrained model finetuning
   • various classification metrics
   • metrics visualizaiton
   • FocalLoss and OneCycle usage examples
   • class imbalance handling
   • model inference
2. tutorial - segmentation
   • car segmentation dataset
   • augmentations with albumentations library
   • training in FP16 with NVIDIA Apex
   • using segmentation models from catalyst/contrib/models/segmentation
   • training with multiple criterion (Dice + IoU + BCE) example
   • Lookahead + RAdam optimizer usage example
   • tensorboard logs visualization
   • predictions visualization
   • Test-time augmentations with ttach library
3. tutorial - table data regression
   • dataset California housing dataset(sklearn)
   • StandardScaler preprocessing
   • Simple MLP (40,10,1) linear layers
   • Training + Inference
   • Results viz.

Usage examples

4. features – classification
   • cifar10 classification model
   • Runner usage example
5. features – segmentation
   • segmentation with unet
   • model training and inference
   • predictions visialization

---

DL pipelines

1. features – model training
   • configuration files usage example
   • local and docker runs
   • metrics visualization with tensorboard
2. features – model training with stages
   • pipeline example with stages
3. features - vanilla GAN on MNIST
   • experiment with multiple phases & models & optimizers
4. tutorial – classification
   • classification model training and inference
   • different augmentations and stages usage
   • knn index model example
   • embeddings projector
   • LrFinder usage
   • grid search metrics visualization
5. tutorial – autolabel
   • pseudolabeling for your dataset
6. [tutorial – segmentation][WIP]
7. [tutorial – autounet][WIP]

---

RL pipelines

1. features – OpenAI Gym LunarLander
   • off-policy RL for continuous action space environment
   • DDPG, SAC, TD3 benchmark
   • async multi-cpu, multi-gpu training
2. features – Atari
   • off-policy RL for discrete action space environment
   • DQN
   • image-based environment with various wrappers
   • CNN-based agent with different distribution heads support

---

Catalyst-info

Link

Contributions

We supervise the Awesome Catalyst list. You can make a PR with your project to the list.