diff --git a/examples/sentiment/scripts/gpt2-sentiment.py b/examples/sentiment/scripts/gpt2-sentiment.py
index 03fad03025..4eab354a68 100644
--- a/examples/sentiment/scripts/gpt2-sentiment.py
+++ b/examples/sentiment/scripts/gpt2-sentiment.py
@@ -20,8 +20,8 @@
 from transformers import pipeline, AutoTokenizer
 from datasets import load_dataset
 
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead
-from trl.core import LengthSampler, set_seed
+from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead, set_seed
+from trl.core import LengthSampler
 
 ########################################################################
 # This is a fully working simple example to use trl with accelerate.
diff --git a/examples/sentiment/scripts/t5-sentiment.py b/examples/sentiment/scripts/t5-sentiment.py
index 1dcf105733..aa88c226b5 100644
--- a/examples/sentiment/scripts/t5-sentiment.py
+++ b/examples/sentiment/scripts/t5-sentiment.py
@@ -19,8 +19,8 @@
 from transformers import pipeline, AutoTokenizer
 from datasets import load_dataset
 
-from trl import PPOTrainer, PPOConfig, AutoModelForSeq2SeqLMWithValueHead
-from trl.core import LengthSampler, set_seed
+from trl import PPOTrainer, PPOConfig, AutoModelForSeq2SeqLMWithValueHead, set_seed
+from trl.core import LengthSampler
 
 ########################################################################
 # This is a fully working simple example to use trl with accelerate.
diff --git a/tests/trainer/test_ppo_trainer.py b/tests/trainer/test_ppo_trainer.py
index 0c03d71168..560d04e89e 100644
--- a/tests/trainer/test_ppo_trainer.py
+++ b/tests/trainer/test_ppo_trainer.py
@@ -97,7 +97,6 @@ def tearDownClass(cls):
                 pass
 
     def setUp(self):
-
         # model_id
         model_id = "gpt2"
 
diff --git a/trl/__init__.py b/trl/__init__.py
index fc6533916a..0c1001d2f7 100644
--- a/trl/__init__.py
+++ b/trl/__init__.py
@@ -2,6 +2,7 @@
 
 __version__ = "0.2.2.dev0"
 
+from .core import set_seed
 from .models import (
     AutoModelForCausalLMWithValueHead,
     AutoModelForSeq2SeqLMWithValueHead,