added a method to create a txt for the ctc decoder

2 files changed, 7 insertions, 0 deletions

diff --git a/swr2_asr/utils/decoder.py b/swr2_asr/utils/decoder.py
index fcddb79..ef8de49 100644
--- a/swr2_asr/utils/decoder.py
+++ b/swr2_asr/utils/decoder.py
@@ -24,3 +24,4 @@ def greedy_decoder(output, labels, label_lengths, tokenizer: CharTokenizer, coll
 
 
 # TODO: add beam search decoder
+
diff --git a/swr2_asr/utils/tokenizer.py b/swr2_asr/utils/tokenizer.py
index 1cc7b84..ee89cdb 100644
--- a/swr2_asr/utils/tokenizer.py
+++ b/swr2_asr/utils/tokenizer.py
@@ -120,3 +120,9 @@ class CharTokenizer:
                     load_tokenizer.char_map[char] = int(index)
                     load_tokenizer.index_map[int(index)] = char
         return load_tokenizer
+    
+    def create_txt(self,path:str):
+      with open(path, 'w',encoding="utf-8") as file:
+        for key,value in self.char_map():
+           file.write(f"{key}\n")
+        
\ No newline at end of file