* Update WER after measuring with final script.

README.md

@@ -23,7 +23,7 @@ model-index:
     metrics:
        - name: Test WER
          type: wer
-         value: 10.61
+         value: 10.50
 ---
 
 # Wav2Vec2-Large-XLSR-53-Spanish
@@ -179,7 +179,7 @@ print("WER: {:2f}".format(100 * chunked_wer(result["sentence"], result["pred_str
 
 ```
 
-**Test Result**: 10.61 %
+**Test Result**: 10.50 %
 
 ## Text processing
 
@@ -198,7 +198,7 @@ For dataset handling reasons, I initially split `train`+`validation` in 10% spli
 * I trained for 30 epochs on the first split only, using similar values as the ones proposed by Patrick in his demo notebook. I used a batch_size of 24 with 2 gradient accumulation steps. This gave a WER of about 16.3%on the full test set.
 * I then trained the resulting model on the 9 remaining splits, for 3 epochs each, but with a faster warmup of 75 steps.
 * Next, I trained 3 epochs on each of the 10 splits using a smaller learning rate of `1e-4`. A warmup of 75 steps was used in this case too. The final model had a WER of about 11.7%.
-* By this time we had already figured out the reason for the initial delay in training time, and I decided to use the full dataset for training. However, in my tests I had seen that varying the learning rate seemed to work well, so I wanted to replicate that. I selected a cosine schedule with hard restarts, a reference learning rate of `3e-5` and 10 epochs. I configured the cosine schedule to have 10 cycles too, and used no warmup. This produced a WER of ~10.6%.
+* By this time we had already figured out the reason for the initial delay in training time, and I decided to use the full dataset for training. However, in my tests I had seen that varying the learning rate seemed to work well, so I wanted to replicate that. I selected a cosine schedule with hard restarts, a reference learning rate of `3e-5` and 10 epochs. I configured the cosine schedule to have 10 cycles too, and used no warmup. This produced a WER of ~10.5%.
 
 
 ## Other things I tried