From 868cdd9708d9e27e63903f54e935c437963e82c9 Mon Sep 17 00:00:00 2001
From: innat <innat.dev@gmail.com>
Date: Wed, 3 Apr 2024 18:04:43 +0600
Subject: [PATCH] update readme

---
 README.md | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index e842652..37176b3 100644
--- a/README.md
+++ b/README.md
@@ -80,6 +80,11 @@ backbone = VideoSwinT(
 )
 ```
 
+Or, we use use the `VideoSwinBackbone` API directly from `from videoswin.backbone`.
+
+
+**Arbitrary Input Shape**
+
 By default, the video swin officially is trained with input shape of `32, 224, 224, 3`. But, We can load the model with different shape. And also load the pretrained weight partially.
 
 ```python
@@ -88,9 +93,10 @@ model = VideoSwinT(
     include_rescaling=False,
     num_classes=10,
 )
-model.load_weights('model.weights.h5', skip_mismatch=True)
+model.load_weights('...weights.h5', skip_mismatch=True)
 ```
 
+
 **Guides**
 
 - To ensure the keras reimplementation with official torch: [logit comparison](guides/video-swin-transformer-keras-and-torchvision.ipynb)