add details in readme

Signed-off-by: Can-Zhao <[email protected]>

Author: Can-Zhao

generation/maisi/README.md

@@ -67,10 +67,9 @@ We retrained several state-of-the-art diffusion model-based methods using our da
 | [512x512x512](./configs/config_infer_80g_512x512x512.json)   |128x128x128| [80,80,80], 8 patches                  | 2                           | 44G         | 569s    | 30s      |
 | [512x512x768](./configs/config_infer_24g_512x512x768.json)   |128x128x192| [80,80,112], 8 patches                 | 4                           | 55G         | 904s    | 48s      |
 
-When `autoencoder_sliding_window_infer_size` is equal or larger than the latent feature size, sliding window will not be used,
-and the time and memory cost remain the same.
+**Table 3:** Inference Time Cost and GPU Memory Usage. `DM Time` refers to the time cost of diffusion model inference. `VAE Time` refers to the time cost of VAE decoder inference. When `autoencoder_sliding_window_infer_size` is equal or larger than the latent feature size, sliding window will not be used,
+and the time and memory cost remain the same. The experiment was tested on A100 80G GPU.
 
-The experiment was tested on A100 80G GPU.
 
 During inference, the peak GPU memory usage happens during the autoencoder decoding latent features.
 To reduce GPU memory usage, we can either increasing `autoencoder_tp_num_splits` or reduce `autoencoder_sliding_window_infer_size`.