diff --git a/lerobot/common/envs/factory.py b/lerobot/common/envs/factory.py
index 72a18c14..770ea392 100644
--- a/lerobot/common/envs/factory.py
+++ b/lerobot/common/envs/factory.py
@@ -11,7 +11,6 @@ def make_env(cfg):
         "from_pixels": cfg.from_pixels,
         "pixels_only": cfg.pixels_only,
         "image_size": cfg.image_size,
-        "max_episode_length": cfg.episode_length,
     }
 
     if cfg.env == "simxarm":
diff --git a/lerobot/common/envs/pusht.py b/lerobot/common/envs/pusht.py
index 02066d31..adc8c015 100644
--- a/lerobot/common/envs/pusht.py
+++ b/lerobot/common/envs/pusht.py
@@ -29,14 +29,12 @@ class PushtEnv(EnvBase):
         image_size=None,
         seed=1337,
         device="cpu",
-        max_episode_length=300,
     ):
         super().__init__(device=device, batch_size=[])
         self.frame_skip = frame_skip
         self.from_pixels = from_pixels
         self.pixels_only = pixels_only
         self.image_size = image_size
-        self.max_episode_length = max_episode_length
 
         if pixels_only:
             assert from_pixels
diff --git a/lerobot/scripts/train.py b/lerobot/scripts/train.py
index 23d92887..6af8656e 100644
--- a/lerobot/scripts/train.py
+++ b/lerobot/scripts/train.py
@@ -80,7 +80,7 @@ def train(cfg: dict, out_dir=None, job_name=None):
             alpha=cfg.per_alpha,
             beta=cfg.per_beta,
             num_slices=num_traj_per_batch,
-            strict_length=False,
+            strict_length=True,
         )
 
         online_buffer = TensorDictReplayBuffer(