From d6dc1d742e95e16e2621e5908e52e8d43dea0784 Mon Sep 17 00:00:00 2001
From: k-for-code <59497032+k-for-code@users.noreply.github.com>
Date: Sat, 23 Jan 2021 17:38:07 +0530
Subject: [PATCH] Update 08.sentiment-analysis-with-bert.ipynb

Bert uses only the encoders from transformers architecture, there was a typo
---
 08.sentiment-analysis-with-bert.ipynb | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/08.sentiment-analysis-with-bert.ipynb b/08.sentiment-analysis-with-bert.ipynb
index d249946..c84f845 100644
--- a/08.sentiment-analysis-with-bert.ipynb
+++ b/08.sentiment-analysis-with-bert.ipynb
@@ -116,7 +116,7 @@
         "\n",
         "*Label* = *NotNext*\n",
         "\n",
-        "The training corpus was comprised of two entries: [Toronto Book Corpus](https://arxiv.org/abs/1506.06724) (800M words) and English Wikipedia (2,500M words). While the original Transformer has an encoder (for reading the input) and a decoder (that makes the prediction), BERT uses only the decoder.\n",
+        "The training corpus was comprised of two entries: [Toronto Book Corpus](https://arxiv.org/abs/1506.06724) (800M words) and English Wikipedia (2,500M words). While the original Transformer has an encoder (for reading the input) and a decoder (that makes the prediction), BERT uses only the encoder.\n",
         "\n",
         "BERT is simply a pre-trained stack of Transformer Encoders. How many Encoders? We have two versions - with 12 (BERT base) and 24 (BERT Large).\n",
         "\n",
@@ -2717,4 +2717,4 @@
       ]
     }
   ]
-}
\ No newline at end of file
+}