diff --git a/init2winit/dataset_lib/pg19.py b/init2winit/dataset_lib/pg19.py
index db2afec5..a5b47fd8 100644
--- a/init2winit/dataset_lib/pg19.py
+++ b/init2winit/dataset_lib/pg19.py
@@ -27,7 +27,7 @@
 
 This module implements a preprocessed PG-19 dataset from TFRecords. The PG-19
 textfiles were tokenized and encoded with SubwordTextEncoder and aggregated into
-tensors of maximum lenght of 8192.
+tensors of maximum length of 8192.
 """
 
 import functools
diff --git a/init2winit/model_lib/adabelief_densenet.py b/init2winit/model_lib/adabelief_densenet.py
index 9b06ac5e..18d20b50 100644
--- a/init2winit/model_lib/adabelief_densenet.py
+++ b/init2winit/model_lib/adabelief_densenet.py
@@ -120,7 +120,7 @@ def __call__(self, x, train):
 class DenseNet(nn.Module):
   """Adabelief DenseNet.
 
-  The network consists of an inital convolutaional layer, four dense blocks
+  The network consists of an initial convolutional layer, four dense blocks
   connected by transition blocks, a pooling layer and a classification layer.
   """
   num_layers: int
diff --git a/init2winit/model_lib/base_model.py b/init2winit/model_lib/base_model.py
index 66538a52..6ffa4085 100644
--- a/init2winit/model_lib/base_model.py
+++ b/init2winit/model_lib/base_model.py
@@ -308,7 +308,7 @@ def _apply_override(sharding, param_type, param_shape):
     return overriden_shardings
 
   def get_sharding(self, params, mesh):
-    """Returns the overriden sharding annotations for the model.
+    """Returns the overridden sharding annotations for the model.
 
     The default sharding strategy is to replicate all layers on all devices.
     Models can override get_sharding_overrides() to specify sharding overrides
diff --git a/init2winit/optimizer_lib/kitchen_sink/_src/transform.py b/init2winit/optimizer_lib/kitchen_sink/_src/transform.py
index db147e14..259a94c5 100644
--- a/init2winit/optimizer_lib/kitchen_sink/_src/transform.py
+++ b/init2winit/optimizer_lib/kitchen_sink/_src/transform.py
@@ -1898,7 +1898,7 @@ def update_fn(updates, state, params):
   return optax.GradientTransformation(init_fn, update_fn)
 
 
-# scale_by_rms exists only for backward compatability
+# scale_by_rms exists only for backward compatibility
 _composites = {
     'scale_by_adaptive_gd': scale_by_adaptive_gd,
     'scale_by_adaptive_gd_simple': scale_by_adaptive_gd_simple,
diff --git a/init2winit/optimizer_lib/samuel.py b/init2winit/optimizer_lib/samuel.py
index 5b6a0f1f..41179e42 100644
--- a/init2winit/optimizer_lib/samuel.py
+++ b/init2winit/optimizer_lib/samuel.py
@@ -64,7 +64,7 @@ def samuel(
     mw_etas: list of multiplicative weight etas.
     seed: initial jax random seed.
     train_loss: train loss to be injected at update time.
-    learning_rate: for compatability, but ignored for now.
+    learning_rate: for compatibility, but ignored for now.
 
   Returns:
     samuel optimizer
diff --git a/init2winit/optimizer_lib/search_subspace.py b/init2winit/optimizer_lib/search_subspace.py
index 58bc33a1..35c9274d 100644
--- a/init2winit/optimizer_lib/search_subspace.py
+++ b/init2winit/optimizer_lib/search_subspace.py
@@ -15,7 +15,7 @@
 
 """Algorithms for narrowing hyperparameter search spaces.
 
-TODO(dsuo): suport discrete hparams.
+TODO(dsuo): support discrete hparams.
 """
 import copy
 import itertools
diff --git a/init2winit/schedules.py b/init2winit/schedules.py
index 8d27c978..a4f55f1e 100644
--- a/init2winit/schedules.py
+++ b/init2winit/schedules.py
@@ -25,7 +25,7 @@
 def _check_schedule_hparams(schedule_hparams, expected_keys):
   if set(schedule_hparams.keys()) != set(expected_keys):
     raise ValueError(
-        'Provided schedule_hparams keys are invalid. Recieved: {}, Expected: {}'
+        'Provided schedule_hparams keys are invalid. Received: {}, Expected: {}'
         .format(sorted(schedule_hparams.keys()), sorted(expected_keys))
     )