From 7a068ec4e15224f5a1e3f23a7bcbd9bd4b030181 Mon Sep 17 00:00:00 2001
From: tf-model-analysis-team <tensorflow-extended-team@google.com>
Date: Thu, 17 Oct 2024 11:34:28 -0700
Subject: [PATCH] Add nltk.download('punkt_tab') to Rouge metric.

This is needed to use the nltk.sent_tokenize() downstream.

PiperOrigin-RevId: 686987036
---
 RELEASE.md                                 | 1 +
 tensorflow_model_analysis/metrics/rouge.py | 1 +
 2 files changed, 2 insertions(+)

diff --git a/RELEASE.md b/RELEASE.md
index ea163a3272..3eb8b48d36 100644
--- a/RELEASE.md
+++ b/RELEASE.md
@@ -19,6 +19,7 @@
 *   Modifies a ROUGE Test to be compatible with NumPy v2.0.1.
 *   Remove keras_util_test.py which is based on estimator models.
 *   Remove dependency on eval_saved_model encodings.
+*   Downloads `punkt_tab` in Rouge metric.
 
 ## Breaking Changes
 
diff --git a/tensorflow_model_analysis/metrics/rouge.py b/tensorflow_model_analysis/metrics/rouge.py
index 45eef52b42..bd5006ff70 100644
--- a/tensorflow_model_analysis/metrics/rouge.py
+++ b/tensorflow_model_analysis/metrics/rouge.py
@@ -93,6 +93,7 @@ def setup(self):
       if not tokenizer_installed:
         logging.info(_LOGGING_MESSAGE_TOKENIZER_PREPARER)
         nltk.download('punkt')
+        nltk.download('punkt_tab')
 
   def create_accumulator(self) -> _Accumulator:
     return _Accumulator()