Update inference.md

video writer need a close() to prevent video length is not enough

documentation/inference.md

@@ -121,6 +121,7 @@ with torch.no_grad():
     for src in DataLoader(reader):
         fgr, pha, *rec = model(src.cuda(), *rec, downsample_ratio=0.25)  # Cycle the recurrent states.
         writer.write(fgr * pha + bgr * (1 - pha))
+    writer.close()
 ```
 
 Or you can use the provided video converter:
@@ -349,4 +350,4 @@ for src in YOUR_VIDEO:  # src is PIL.Image.
 
 ```
 
-Our CoreML models only support fixed resolutions. If you need other resolutions, you can export them yourself. See [coreml branch](https://github.com/PeterL1n/RobustVideoMatting/tree/coreml) for model export. 
+Our CoreML models only support fixed resolutions. If you need other resolutions, you can export them yourself. See [coreml branch](https://github.com/PeterL1n/RobustVideoMatting/tree/coreml) for model export.