diff --git a/opensora/models/diffusion/dit/dit.py b/opensora/models/diffusion/dit/dit.py
index 676f2cd6a..d7611b2b5 100644
--- a/opensora/models/diffusion/dit/dit.py
+++ b/opensora/models/diffusion/dit/dit.py
@@ -267,7 +267,8 @@ class DiT(nn.Module):
         nn.init.constant_(self.x_embedder.proj.bias, 0)
 
         # Initialize label embedding table:
-        nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)
+        if self.extras == 2:
+            nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)
 
         # Initialize timestep embedding MLP:
         nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
@@ -308,7 +309,8 @@ class DiT(nn.Module):
         attention_mask = attention_mask.flatten(1).unsqueeze(-1)  # bs t h w -> bs thw 1
         attention_mask = attention_mask @ attention_mask.transpose(1, 2)  # bs thw 1 @ bs 1 thw = bs thw thw
         attention_mask = attention_mask.unsqueeze(1)
-        attention_mask = attention_mask.masked_fill(attention_mask == 0, torch.finfo(dtype).min)
+        # attention_mask = attention_mask.masked_fill(attention_mask == 0, torch.finfo(dtype).min)
+        attention_mask = attention_mask.masked_fill(attention_mask == 0, 1e-8)
         return attention_mask
 
     def forward(self, x, t, y, attention_mask):
diff --git a/opensora/sample/placeholder b/opensora/sample/placeholder
deleted file mode 100644
index e69de29bb..000000000
diff --git a/opensora/train/train.py b/opensora/train/train.py
index d885d1236..f027a04cf 100644
--- a/opensora/train/train.py
+++ b/opensora/train/train.py
@@ -249,6 +249,7 @@ def main(args):
                 steps_per_sec = log_steps / (end_time - start_time)
                 # Reduce loss history over all processes:
                 avg_loss = torch.tensor(running_loss / log_steps, device=device)
+                dist.all_reduce(avg_loss, op=dist.ReduceOp.SUM)
                 avg_loss = avg_loss.item() / accelerator.num_processes
                 # logger.info(f"(step={train_steps:07d}) Train Loss: {avg_loss:.4f}, Train Steps/Sec: {steps_per_sec:.2f}")
                 if accelerator.is_main_process:
diff --git a/opensora/utils/movie2clip.py b/opensora/utils/movie2clip.py
deleted file mode 100644
index 2da7a1d18..000000000
--- a/opensora/utils/movie2clip.py
+++ /dev/null
@@ -1,66 +0,0 @@
-import cv2
-from tqdm import tqdm
-
-
-def crop_video(input_file, start_and_end):
-    # 打开视频文件
-    video = cv2.VideoCapture(input_file)
-
-    # 获取视频的帧速率和总帧数
-    fps = video.get(cv2.CAP_PROP_FPS)
-    total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
-
-
-    current_frame = 0
-    # 设置当前帧的索引
-    for s, e in tqdm(start_and_end):
-        # 计算裁剪的起始帧和结束帧
-        start_frame = int(s * fps)
-        end_frame = int(e * fps)
-        # 设置视频编解码器并创建输出视频对象
-        fourcc = cv2.VideoWriter_fourcc(*'mp4v')
-        output = cv2.VideoWriter(f'clip/{s}-{e}.mp4', fourcc, fps, (int(video.get(3)), int(video.get(4))))
-
-
-        # 读取并写入视频帧，直到达到结束帧
-        while current_frame < end_frame:
-            ret, frame = video.read()
-            if not ret:
-                break
-            if current_frame >= start_frame:
-                output.write(frame)
-            current_frame += 1
-
-        # 释放视频对象和输出对象
-    video.release()
-    output.release()
-
-    print("视频裁剪完成！")
-
-# 使用示例
-input_file = r"C:\Users\ABin\Desktop\S01E03 热带宁静Tropical Serenity_超清 4K.mp4"
-start_and_end = [
-    [24, 71],
-    [76, 178],
-    [182, 265],
-    [268, 315],
-    [322, 366],
-    [372, 467],
-    [473, 535],
-    [550, 610],
-    [618, 684],
-    [689, 792],
-    [800, 880],
-    [884, 945],
-    [951, 1002],
-    [1009, 1116],
-    [1127, 1196],
-    [1205, 1301],
-    [1306, 1381],
-    [1386, 1421],
-    [1430, 1505],
-    [1571, 1628],
-    [1636, 1724],
-    [1731, 1784],
-]
-crop_video(input_file, start_and_end)
\ No newline at end of file