fix: typo (explanation) (#167)

docs/examples/config.rst

 .. _config-explain-page:
 
-Config Explaination
+Config Explanation
 ===================
 
 ppo_trainer.yaml for FSDP Backend

docs/examples/gsm8k_example.rst

@@ -101,7 +101,7 @@ Step 4: Perform PPO training with your model on GSM8K Dataset
 - Users could replace the ``data.train_files`` ,\ ``data.val_files``,
   ``actor_rollout_ref.model.path`` and ``critic.model.path`` based on
   their environment.
-- See :doc:`config` for detailed explaination of each config field.
+- See :doc:`config` for detailed explanation of each config field.
 
 **Reward Model/Function**
 

docs/start/quickstart.rst

@@ -136,7 +136,7 @@ If you encounter out of memory issues with HBM less than 32GB, enable the follow
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=1 \
     critic.ppo_micro_batch_size_per_gpu=1 \
 
-For the full set of configs, please refer to :ref:`config-explain-page` for detailed explaination and performance tuning.
+For the full set of configs, please refer to :ref:`config-explain-page` for detailed explanation and performance tuning.
 
 
 .. [1] The original paper (https://arxiv.org/pdf/2110.14168) mainly focuses on training a verifier (a reward model) to solve math problems via Best-of-N sampling. In this example, we train an RL agent using a rule-based reward model.