New project Quick Start
To start a new project define two files, a LightningModule and a Trainer file.
To illustrate Lightning power and simplicity, here's an example of a typical research flow.
Case 1: BERT
Let's say you're working on something like BERT but want to try different ways of training or even different networks.
You would define a single LightningModule and use flags to switch between your different ideas.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Case 2: COOLER NOT BERT
But if you wanted to try something completely different, you'd define a new module for that.
1 2 3 4 5 6 7
Rapid research flow
Then you could do rapid research by switching between these two and using the same trainer.
1 2 3 4 5 6 7
Notice a few things about this flow:
1. You're writing pure PyTorch... no unnecessary abstractions or new libraries to learn.
2. You get free GPU and 16-bit support without writing any of that code in your model.
3. You also get all of the capabilities below (without coding or testing yourself).
Quick start examples
- CPU example
- Hyperparameter search on single GPU
- Hyperparameter search on multiple GPUs on same node
- Hyperparameter search on a SLURM HPC cluster
Computing cluster (SLURM)
- Fast dev run
- Inspect gradient norms
- Log GPU usage
- Make model overfit on subset of data
- Print the parameter count by layer
- Pring which gradients are nan
- Print input and output size of every module in system
- Implement Your Own Distributed (DDP) training
- 16-bit mixed precision
- Single GPU
- Self-balancing architecture
- Display metrics in progress bar
- Log metric row every k batches
- Process position
- Tensorboard support
- Save a snapshot of all hyperparameters
- Snapshot code for a training run
- Write logs file to csv every k batches
- Accumulate gradients
- Force training for min or max epochs
- Early stopping callback
- Force disable early stop
- Gradient Clipping
- Learning rate scheduling
- Use multiple optimizers (like GANs)
- Set how much of the training set to check (1-100%)
- Step optimizers at arbitrary intervals
- Check validation every n epochs
- Set how much of the validation set to check
- Set how much of the test set to check
- Set validation check frequency within 1 training epoch
- Set the number of validation sanity steps