Training in progress, step 1600

.ipynb_checkpoints/run_speech_recognition_ctc-checkpoint.py

@@ -439,17 +439,16 @@ def main():
 
     def remove_special_characters(batch):
         if chars_to_ignore_regex is not None:
-            batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
+            batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).lower() + " "
+            batch["sentence"] = re.sub('!', '', batch["sentence"]).lower() + " "
+            batch["sentence"] = batch["sentence"].replace('â', 'a').replace('á', 'a').replace('ō', 'o').lower() + " "
+            batch["sentence"] = batch["sentence"].replace('û', 'u').replace('ł', 'l').replace('é', 'e').lower() + " "
+            batch["sentence"] = batch["sentence"].replace('ń',"n").replace('î',"").replace('ô',"").replace('û',"").lower() + " "
+            batch["sentence"] = batch["sentence"].replace('\"',"").replace("&","").replace("'","").replace("(","").lower() + " "
+            batch["sentence"] = batch["sentence"].replace('[',"").replace("]","").replace("\\","").replace(")","").lower() + " "
+            batch["sentence"] = batch["sentence"].replace("  "," ").replace("  "," ").replace("  "," ").lower() + " "
         else:
-            batch["target_text"] = batch[text_column_name].lower() + " "
-            batch["target_text"] = re.sub('[â]', 'a', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[î]', 'i', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[ô]', 'o', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[û]', 'u', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('!', '', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[û]', 'u', batch[text_column_name]).lower() + " "
-            batch["target_text"] = batch[text_column_name].replace('\"',"").replace("&","").replace("'","").replace("(","").lower() + " "
-            batch["target_text"] = batch[text_column_name].replace('[',"").replace("]","").replace("\\","").replace(")","").lower() + " "
+            batch["sentence"] = batch["sentence"].lower() + " "
         return batch
 
     with training_args.main_process_first(desc="dataset map special characters removal"):

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b491618ba7f98aa5aaa423d998ee9b9ba88ee3b6e91a5f322e641d6e22f6f33d
+oid sha256:a2da86a14f166346a4c719264af09e51aa06466d45a74e9e6ecc06aea1d14528
 size 1262091761

run_speech_recognition_ctc.py

@@ -439,17 +439,16 @@ def main():
 
     def remove_special_characters(batch):
         if chars_to_ignore_regex is not None:
-            batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
+            batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).lower() + " "
+            batch["sentence"] = re.sub('!', '', batch["sentence"]).lower() + " "
+            batch["sentence"] = batch["sentence"].replace('â', 'a').replace('á', 'a').replace('ō', 'o').lower() + " "
+            batch["sentence"] = batch["sentence"].replace('û', 'u').replace('ł', 'l').replace('é', 'e').lower() + " "
+            batch["sentence"] = batch["sentence"].replace('ń',"n").replace('î',"").replace('ô',"").replace('û',"").lower() + " "
+            batch["sentence"] = batch["sentence"].replace('\"',"").replace("&","").replace("'","").replace("(","").lower() + " "
+            batch["sentence"] = batch["sentence"].replace('[',"").replace("]","").replace("\\","").replace(")","").lower() + " "
+            batch["sentence"] = batch["sentence"].replace("  "," ").replace("  "," ").replace("  "," ").lower() + " "
         else:
-            batch["target_text"] = batch[text_column_name].lower() + " "
-            batch["target_text"] = re.sub('[â]', 'a', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[î]', 'i', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[ô]', 'o', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[û]', 'u', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('!', '', batch[text_column_name]).lower() + " "
-            batch["target_text"] = re.sub('[û]', 'u', batch[text_column_name]).lower() + " "
-            batch["target_text"] = batch[text_column_name].replace('\"',"").replace("&","").replace("'","").replace("(","").lower() + " "
-            batch["target_text"] = batch[text_column_name].replace('[',"").replace("]","").replace("\\","").replace(")","").lower() + " "
+            batch["sentence"] = batch["sentence"].lower() + " "
         return batch
 
     with training_args.main_process_first(desc="dataset map special characters removal"):

wandb/run-20220130_033240-1ciqu2rb/files/output.log

@@ -2288,3 +2288,765 @@ Configuration saved in ./checkpoint-1200/config.json
 {'eval_loss': 0.3779752850532532, 'eval_wer': 0.3348087406233984, 'eval_runtime': 176.2153, 'eval_samples_per_second': 20.475, 'eval_steps_per_second': 2.559, 'epoch': 9.3}
 Model weights saved in ./checkpoint-1200/pytorch_model.bin
 Configuration saved in ./checkpoint-1200/preprocessor_config.json
+Configuration saved in ./preprocessor_config.json
+Deleting older checkpoint [checkpoint-400] due to args.save_total_limit
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ 34%|█████████████████████████████████████▉                                                                           | 1300/3870 [2:02:31<3:12:24,  4.49s/it]The following columns in the evaluation set  don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.
+***** Running Evaluation *****
+  Num examples = 3608
+  Batch size = 8
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+100%|█████████████████████████████████████████████████████████████████████████████████████████████████████| 451/451 [02:50<00:00,  2.53it/s]
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ 36%|████████████████████████████████████████▉                                                                        | 1400/3870 [2:11:49<2:48:02,  4.08s/it]The following columns in the evaluation set  don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.
+***** Running Evaluation *****
+  Num examples = 3608
+  Batch size = 8
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+100%|█████████████████████████████████████████████████████████████████████████████████████████████████████| 451/451 [02:51<00:00,  2.53it/s]
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ 39%|███████████████████████████████████████████▊                                                                     | 1499/3870 [2:21:03<2:35:27,  3.93s/it]
+ 39%|███████████████████████████████████████████▊                                                                     | 1500/3870 [2:21:07<2:34:01,  3.90s/it]The following columns in the evaluation set  don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.
+***** Running Evaluation *****
+  Num examples = 3608
+  Batch size = 8
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+100%|█████████████████████████████████████████████████████████████████████████████████████████████████████| 451/451 [02:51<00:00,  2.41it/s]
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ 41%|██████████████████████████████████████████████▋                                                                  | 1600/3870 [2:30:27<2:17:07,  3.62s/it]The following columns in the evaluation set  don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.
+***** Running Evaluation *****
+  Num examples = 3608
+  Batch size = 8
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Configuration saved in ./checkpoint-1600/config.json
+{'eval_loss': 0.3882218301296234, 'eval_wer': 0.3162186087685785, 'eval_runtime': 175.3702, 'eval_samples_per_second': 20.574, 'eval_steps_per_second': 2.572, 'epoch': 12.4}
+Model weights saved in ./checkpoint-1600/pytorch_model.bin
+Configuration saved in ./checkpoint-1600/preprocessor_config.json