fix normalization and data type

lerobot/scripts/visualize_dataset.py

@@ -25,8 +25,10 @@ def visualize_dataset_cli(cfg: dict):
 
 
 def cat_and_write_video(video_path, frames, fps):
+    # Expects images in [0, 1].
     frames = torch.cat(frames)
-    assert frames.dtype == torch.uint8
+    assert frames.max() <= 1 and frames.min() >= 0
+    frames = (255 * frames).to(dtype=torch.uint8)
     frames = einops.rearrange(frames, "b c h w -> b h w c").numpy()
     imageio.mimsave(video_path, frames, fps=fps)
 
@@ -59,7 +61,7 @@ def visualize_dataset(cfg: dict, out_dir=None):
         ep_td = offline_buffer.sample(1)
         ep_idx = ep_td["episode"][FIRST_FRAME].item()
 
-        # TODO(rcaene): modify offline_buffer._sampler._sample_list or sampler to randomly sample an episode, but sequentially sample frames
+        # TODO(rcadene): modify offline_buffer._sampler._sample_list or sampler to randomly sample an episode, but sequentially sample frames
         no_more_frames = offline_buffer._sampler._sample_list.numel() == 0
         new_episode = ep_idx != current_ep_idx
 
@@ -69,7 +71,7 @@ def visualize_dataset(cfg: dict, out_dir=None):
         for im_key in offline_buffer.image_keys:
             if new_episode or no_more_frames:
                 # append last observed frames (the ones after last action taken)
-                frames[im_key].append(ep_td[("next", *im_key)])
+                frames[im_key].append(offline_buffer.transform(ep_td["next"])[im_key])
 
                 video_dir = Path(out_dir) / "visualize_dataset"
                 video_dir.mkdir(parents=True, exist_ok=True)
@@ -101,6 +103,9 @@ def visualize_dataset(cfg: dict, out_dir=None):
             logging.info("Ran out of frames")
             break
 
+        if current_ep_idx == NUM_EPISODES_TO_RENDER:
+            break
+
     for thread in threads:
         thread.join()