From 1b342259b457ba7d9367eb3f6ee40e7b8485ce3a Mon Sep 17 00:00:00 2001 From: Vahid Zehtab <33608325+vahidzee@users.noreply.github.com> Date: Thu, 13 Apr 2023 19:56:12 -0400 Subject: [PATCH] update general docs links --- README.md | 2 +- docs/README.md | 64 -------------------------------------------------- 2 files changed, 1 insertion(+), 65 deletions(-) delete mode 100644 docs/README.md diff --git a/README.md b/README.md index dd1f0b5..f789bd8 100644 --- a/README.md +++ b/README.md @@ -1,7 +1,7 @@ # LightningToolbox: A PyTorch Lightning Facilitator

Installation • - Docs • + Docs • License

diff --git a/docs/README.md b/docs/README.md deleted file mode 100644 index abbf5ad..0000000 --- a/docs/README.md +++ /dev/null @@ -1,64 +0,0 @@ - -## Usage -### TrainingModule & DataModule - -```python -from lightning_toolbox import TrainingModule, DataModule -from lightning import Trainer - -training_module = TrainingModule( - # your model - model = model, # or alternatively provide a model_cls and model_args instead as follows: - # model_cls = `your_model_cls`, or an import path such as `torchvision.models.resnet18` - # model_args = `dict` of arguments to pass to your model_cls constructor - # (e.g. `pretrained=True`, for the case of resnet18) - - # objective function - objective = ... # your objective function, factors, etc. (see the docs for more details) - - # optimizer (or a list of optimizers if you want to use different algorithms for - # different parts of your model) - optimizer = ... # your optimizer (e.g. "torch.optim.AdamW"), you can provide - # optimizer_args to pass to the optimizer constructor - lr = ... # learning rate (e.g. 1e-3) - optimizer_args = ... # optimizer arguments (e.g. `dict(weight_decay=1e-2)`) - optimizer_parameters = ... # parameters to optimize (default is "model.parameters()") - optimizer_frequency = ... # how often to update the optimizer (e.g. 1) - optimizer_is_active = ... # whether the optimizer is active (e.g. True, or a - #function that returns a bool) - # lr scheduler (or a list of lr schedulers) - lr_scheduler = ... # your lr scheduler (e.g. "torch.optim.lr_scheduler.CosineAnnealingLR") - lr_scheduler_args = ... # lr scheduler arguments (e.g. `dict(T_max=10)`) - lr_scheduler_interval = ... # step/epoch - lr_scheduler_monitor = ... # metric to monitor (e.g. "loss/val") - lr_scheduler_frequency = ... # how often to update the lr (e.g. 1) - lr_scheduler_optimizer = ... # the index of the optimizer this scheduler is for - lr_scheduler_name = ... # name of the lr scheduler (e.g. "cosine") - - ... # other arguments (see the docs for more details) -) - -data_module = DataModule( - dataset = ... # your dataset, or an import path - # (e.g. "torchvision.datasets.CIFAR10") - dataset_args = ... # optionally dataset arguments - # (e.g. `dict(root="data", train=True, download=True)`) - - val_size = ... # validation set size - # (e.g. 0.2 of the provided dataset, or 1000 samples from it) - # val_dataset = ... # validation dataset - # (e.g. "torchvision.datasets.MNIST") to use instead of val_size - # val_dataset_args = ... # validation dataset arguments - # (e.g. `dict(root="data", train=False, download=True)`) - - batch_size = ... # batch size (e.g. 32) - num_workers = ... # number of workers (e.g. 4) - pin_memory = ... # whether to use pinned memory (e.g. True) - train_shuffle = ... # whether to shuffle the training set (e.g. True) - val_shuffle = ... # whether to shuffle the validation set (e.g. False) - ... # other arguments (see the docs for more details) -) - -trainer = Trainer() -trainer.fit(training_module) -``` \ No newline at end of file