fixes test issues on ddp (#1017)

* updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs * updated checkpoint docs
2026-09-09 11:32:07 +08:00 · 2020-03-02 21:50:38 -05:00
parent 5458d05cd8
commit 6dae5698ef
6 changed files with 72 additions and 1 deletions
@@ -5,6 +5,19 @@ Lightning can automate saving and loading checkpoints.

 Checkpoint saving
 -----------------
+A Lightning checkpoint has everything needed to restore a training session including:
+
+- 16-bit scaling factor (apex)
+- Current epoch
+- Global step
+- Model state_dict
+- State of all optimizers
+- State of all learningRate schedulers
+- State of all callbacks
+- The hyperparameters used for that model if passed in as hparams (Argparse.Namespace)
+
+Automatic saving
+^^^^^^^^^^^^^^^^

 Checkpointing is enabled by default to the current working directory.
 To change the checkpoint path pass in:
@@ -59,6 +72,15 @@ The Lightning checkpoint also saves the hparams (hyperparams) passed into the Li
       def __init__(self, hparams, ...):
           self.hparams = hparams

+Manual saving
+^^^^^^^^^^^^^
+
+To save your own checkpoint call:
+
+.. code-block:: python
+
+   model.save_checkpoint(PATH)
+
 Checkpoint Loading
 ------------------