From 3538c25cc5d660648c6e50d5873289a2e73c450f Mon Sep 17 00:00:00 2001
From: Jordi Mas <jmas@softcatala.org>
Date: Tue, 7 May 2024 13:28:43 +0200
Subject: [PATCH] Update to whisper-ctranslate2 0.4.3 and the depending stack.
 Improvements in WER

---
 benchmark/results.json             | 46 +++++++++++++++---------------
 transcribe-batch/docker/Dockerfile |  2 +-
 transcribe-batch/requirements.txt  |  6 ++--
 3 files changed, 27 insertions(+), 27 deletions(-)

diff --git a/benchmark/results.json b/benchmark/results.json
index 9b9f547..a0695a5 100644
--- a/benchmark/results.json
+++ b/benchmark/results.json
@@ -4,30 +4,30 @@
             {
                 "audio": "samples/15GdH9-curt.mp3",
                 "wer": "72.57",
-                "time": 86
+                "time": 83
             },
             {
                 "audio": "samples/EloiBadiaCat.mp3",
-                "wer": "22.26",
-                "time": 69
+                "wer": "22.57",
+                "time": 63
             },
             {
                 "audio": "samples/Son_Goku_catalan_valencian_voice.ogg",
-                "wer": "51.90",
-                "time": 32
+                "wer": "40.51",
+                "time": 27
             },
             {
                 "audio": "samples/Universal_Declaration_of_Human_Rights_-_cat_-_nv.ogg",
-                "wer": "39.51",
-                "time": 202
+                "wer": "31.38",
+                "time": 200
             },
             {
                 "audio": "samples/Ona_catalan-balear.mp3",
-                "wer": "46.81",
-                "time": 37
+                "wer": "46.01",
+                "time": 38
             },
             {
-                "avg_wer": "46.61"
+                "avg_wer": "42.61"
             }
         ]
     },
@@ -35,38 +35,38 @@
         "medium": [
             {
                 "audio": "samples/15GdH9-curt.mp3",
-                "wer": "35.43",
-                "time": 174
+                "wer": "35.62",
+                "time": 181
             },
             {
                 "audio": "samples/EloiBadiaCat.mp3",
-                "wer": "15.05",
-                "time": 167
+                "wer": "14.73",
+                "time": 175
             },
             {
                 "audio": "samples/Son_Goku_catalan_valencian_voice.ogg",
-                "wer": "25.32",
-                "time": 50
+                "wer": "27.85",
+                "time": 52
             },
             {
                 "audio": "samples/Universal_Declaration_of_Human_Rights_-_cat_-_nv.ogg",
-                "wer": "36.50",
-                "time": 513
+                "wer": "19.12",
+                "time": 532
             },
             {
                 "audio": "samples/Ona_catalan-balear.mp3",
-                "wer": "22.87",
-                "time": 88
+                "wer": "21.54",
+                "time": 92
             },
             {
-                "avg_wer": "27.03"
+                "avg_wer": "23.77"
             }
         ]
     },
     {
         "totals": {
-            "wer": "36.82",
-            "time": "1418"
+            "wer": "33.19",
+            "time": "1443"
         }
     }
 ]
\ No newline at end of file
diff --git a/transcribe-batch/docker/Dockerfile b/transcribe-batch/docker/Dockerfile
index e4731e3..64f96df 100644
--- a/transcribe-batch/docker/Dockerfile
+++ b/transcribe-batch/docker/Dockerfile
@@ -13,7 +13,7 @@ RUN apt-get install python3-pip python3-dev -y --no-install-recommends
 # Solve errors:
 # - Error: libcudnn_ops_infer.so.8: cannot open shared object file: No such file or directory
 # - Error: libcublasLt.so.11: cannot open shared object file: No such file or directory
-RUN apt-get install libcudnn8=8.8.1.3-1+cuda11.8 &&  apt-get install libcublas-11-8 -y
+RUN apt-get install libcudnn8=8.8.1.3-1+cuda11.8 && apt-get install libcublas-12-3 &&  apt-get install libcublas-11-8 -y
 
 WORKDIR /srv
 
diff --git a/transcribe-batch/requirements.txt b/transcribe-batch/requirements.txt
index f4f96a1..561f3a1 100644
--- a/transcribe-batch/requirements.txt
+++ b/transcribe-batch/requirements.txt
@@ -1,5 +1,5 @@
 psutil
-ctranslate2==3.22
-faster-whisper==0.10.1
-whisper-ctranslate2==0.3.7
+ctranslate2==4.2.1
+faster-whisper==1.0.2
+whisper-ctranslate2==0.4.3
 langdetect==1.0.9