Add scripts and weights

run_speech_recognition_whisper.py

@@ -23,6 +23,7 @@ import os
 import whisper
 import sys
 from dataclasses import dataclass, field
+import tempfile
 
 from typing import Optional, Dict, Union, List
 
@@ -275,7 +276,6 @@ class WhisperDataCollatorWithPadding:
     """
 
     eos_token_id: int
-    time_stamp_token_id: int
 
     def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
         """
@@ -626,9 +626,7 @@ def main():
             torch.save(self.args, os.path.join(output_dir, "training_args.bin"))
 
     # Define data collator
-    eos = tokenizer.eos_token_id
-    t_stamp = tokenizer("<|notimestamps|>").input_ids[0]
-    whisper_data_collator = WhisperDataCollatorWithPadding(eos_token_id=eos, time_stamp_token_id=t_stamp)
+    whisper_data_collator = WhisperDataCollatorWithPadding(eos_token_id=tokenizer.eos_token_id)
 
     # make sure model uses 50257 as BOS
     bos = tokenizer("<|startoftranscript|>").input_ids[0]