From e7f1614b4e9cd6a650dbcc100f29c980b9020a4a Mon Sep 17 00:00:00 2001
From: Alejandro Casanova
 <36633619+Alejandro-Casanova@users.noreply.github.com>
Date: Thu, 28 Sep 2023 16:40:24 +0200
Subject: [PATCH] fixed typo

I think??
---
 08_pytorch_paper_replicating.ipynb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/08_pytorch_paper_replicating.ipynb b/08_pytorch_paper_replicating.ipynb
index e3d1ccf1..0b7d48e8 100644
--- a/08_pytorch_paper_replicating.ipynb
+++ b/08_pytorch_paper_replicating.ipynb
@@ -4361,7 +4361,7 @@
     "4. Try repeating the steps from excercise 3 but this time use the \"`ViT_B_16_Weights.IMAGENET1K_SWAG_E2E_V1`\" pretrained weights from [`torchvision.models.vit_b_16()`](https://pytorch.org/vision/stable/models/generated/torchvision.models.vit_b_16.html#torchvision.models.vit_b_16).\n",
     "    * **Note:** ViT pretrained with SWAG weights has a minimum input image size of `(384, 384)` (the pretrained ViT in exercise 3 has a minimum input size of `(224, 224)`), though this is accessible in the weights `.transforms()` method.\n",
     "5. Our custom ViT model architecture closely mimics that of the ViT paper, however, our training recipe misses a few things. Research some of the following topics from Table 3 in the ViT paper that we miss and write a sentence about each and how it might help with training:\n",
-    "    * ImageNet-22k pretraining (more data).\n",
+    "    * ImageNet-21k pretraining (more data).\n",
     "    * Learning rate warmup.\n",
     "    * Learning rate decay.\n",
     "    * Gradient clipping."