updates model

README.md

@@ -25,10 +25,10 @@ model-index:
     metrics:
        - name: Test WER
          type: wer
-         value: 15.345730
+         value: 15.156165
        - name: Test CER
          type: cer
-         value: 5.573185
+         value: 5.343999
 ---
 
 # Wav2Vec2-Large-XLSR-53-Swedish
@@ -127,8 +127,9 @@ print("CER: {:2f}".format(100 * wer.compute(predictions=[" ".join(list(entry)) f
 
 ```
 
-**Test Result**:  15.345730 %
+**Test Result**:  15.156165 %
 
+## Training
 
 As first step used Common Voice train dataset and parts from NST
 as can be found [here](https://github.com/se-asr/nst/tree/master).
@@ -144,4 +145,4 @@ After training like this for 20000 steps the model was finetuned on all of nst d
 mask = [(1 < len(x.split()) < 25) and np.average([len(entry) for entry in x.split()]) > 3 for x in dataset['transcript'].tolist()]
 ```
 
-and all of common voice for 50000 more steps.
+and all of common voice for 100000 more steps approximately 16 epochs.

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48cc1cd2bba0dc6c727049576a354799ba7bdc644ff6a0ac6e06e99d61856659
+oid sha256:b3c27ed20a605dc14d8ff2bccb168d08d947d19902b9fe56871eb89071d5418d
 size 1262065047