Update README.md

README.md

@@ -1,116 +1,33 @@
 ---
-license: apache-2.0
-tags:
-- generated_from_trainer
+language:
+- da
+license: other
+datasets:
+- ftspeech
 model-index:
 - name: wav2vec2-xls-r-300m-ftspeech
   results: []
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
-
-# wav2vec2-xls-r-300m-ftspeech
-
-This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 17.8348
-- Wer: 0.1186
+# XLS-R-300m-FTSpeech
 
 ## Model description
 
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
+This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the [FTSpeech dataset](https://ftspeech.github.io/), being a dataset of 1,800 hours of transcribed speeches from the Danish parliament.
 
-## Training procedure
 
-### Training hyperparameters
+## Performance
 
-The following hyperparameters were used during training:
-- learning_rate: 0.0001
-- train_batch_size: 1
-- eval_batch_size: 1
-- seed: 4242
-- gradient_accumulation_steps: 32
-- total_train_batch_size: 32
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 2000
-- num_epochs: 3
-- mixed_precision_training: Native AMP
+The model achieves the following WER scores (lower is better):
 
-### Training results
+| **Dataset** | **WER without LM** | **WER with 5-gram LM** |
+| :---:   | ---: | ---: |
+| [Danish part of Common Voice 8.0](https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0/viewer/da/train) | 20.48 | 17.91 |
+| [Alvenir test set](https://huggingface.co/datasets/Alvenir/alvenir_asr_da_eval) | 15.46 | 13.84 |
 
-| Training Loss | Epoch | Step  | Validation Loss | Wer    |
-|:-------------:|:-----:|:-----:|:---------------:|:------:|
-| 311.274       | 0.02  | 500   | 329.9529        | 1.0    |
-| 296.713       | 0.03  | 1000  | 305.5616        | 1.0000 |
-| 69.6128       | 0.05  | 1500  | 77.2267         | 0.6466 |
-| 47.4542       | 0.06  | 2000  | 56.0227         | 0.5101 |
-| 39.2415       | 0.08  | 2500  | 40.1751         | 0.3483 |
-| 35.9888       | 0.1   | 3000  | 33.1659         | 0.2619 |
-| 34.1621       | 0.11  | 3500  | 30.4220         | 0.2296 |
-| 32.3383       | 0.13  | 4000  | 28.3836         | 0.2214 |
-| 31.1862       | 0.14  | 4500  | 28.7228         | 0.2220 |
-| 29.818        | 0.16  | 5000  | 28.3220         | 0.2259 |
-| 29.4729       | 0.18  | 5500  | 26.5646         | 0.2024 |
-| 27.6171       | 0.19  | 6000  | 26.3382         | 0.1995 |
-| 27.4549       | 0.21  | 6500  | 24.1257         | 0.1697 |
-| 27.9176       | 0.22  | 7000  | 24.8758         | 0.1945 |
-| 27.4036       | 0.24  | 7500  | 24.1006         | 0.1746 |
-| 26.5633       | 0.26  | 8000  | 23.0034         | 0.1582 |
-| 26.3558       | 0.27  | 8500  | 24.7499         | 0.1913 |
-| 25.9604       | 0.29  | 9000  | 22.5813         | 0.1674 |
-| 25.6154       | 0.31  | 9500  | 22.4642         | 0.1499 |
-| 25.6231       | 0.32  | 10000 | 21.8089         | 0.1534 |
-| 26.7554       | 0.34  | 10500 | 21.9619         | 0.1543 |
-| 25.2901       | 0.35  | 11000 | 22.0643         | 0.1593 |
-| 24.8642       | 0.37  | 11500 | 21.1113         | 0.1480 |
-| 25.4664       | 0.39  | 12000 | 21.2492         | 0.1458 |
-| 24.6433       | 0.4   | 12500 | 20.7650         | 0.1419 |
-| 24.8455       | 0.42  | 13000 | 21.8535         | 0.1490 |
-| 25.1176       | 0.43  | 13500 | 20.7491         | 0.1429 |
-| 24.4585       | 0.45  | 14000 | 20.7948         | 0.1423 |
-| 24.1613       | 0.47  | 14500 | 20.5817         | 0.1431 |
-| 23.7281       | 0.48  | 15000 | 20.1209         | 0.1333 |
-| 23.0396       | 0.5   | 15500 | 20.2883         | 0.1383 |
-| 24.7056       | 0.51  | 16000 | 19.6813         | 0.1330 |
-| 23.608        | 0.53  | 16500 | 20.0252         | 0.1394 |
-| 23.9536       | 0.55  | 17000 | 19.9039         | 0.1341 |
-| 23.1848       | 0.56  | 17500 | 19.9114         | 0.1308 |
-| 23.1835       | 0.58  | 18000 | 19.7044         | 0.1345 |
-| 23.9372       | 0.59  | 18500 | 19.2201         | 0.1296 |
-| 23.2182       | 0.61  | 19000 | 19.3723         | 0.1350 |
-| 22.3118       | 0.63  | 19500 | 19.2624         | 0.1344 |
-| 22.9372       | 0.64  | 20000 | 19.5823         | 0.1387 |
-| 23.1536       | 0.66  | 20500 | 18.9077         | 0.1289 |
-| 22.3477       | 0.67  | 21000 | 18.7098         | 0.1257 |
-| 22.3701       | 0.69  | 21500 | 19.0815         | 0.1300 |
-| 22.6709       | 0.71  | 22000 | 18.4433         | 0.1242 |
-| 22.2519       | 0.72  | 22500 | 18.7482         | 0.1275 |
-| 21.8536       | 0.74  | 23000 | 18.6565         | 0.1236 |
-| 22.4479       | 0.76  | 23500 | 18.6478         | 0.1264 |
-| 21.6824       | 0.77  | 24000 | 18.4383         | 0.1257 |
-| 22.1622       | 0.79  | 24500 | 18.4086         | 0.1212 |
-| 22.2626       | 0.8   | 25000 | 18.4613         | 0.1230 |
-| 21.0009       | 0.82  | 25500 | 18.1851         | 0.1165 |
-| 20.554        | 0.84  | 26000 | 17.7352         | 0.1165 |
-| 21.5141       | 0.85  | 26500 | 18.3084         | 0.1207 |
-| 20.5925       | 0.87  | 27000 | 17.9997         | 0.1207 |
-| 21.0997       | 0.88  | 27500 | 17.7534         | 0.1193 |
-| 21.7098       | 0.9   | 28000 | 17.8348         | 0.1186 |
 
+## Additional Information
 
-### Framework versions
+### Licensing Information
 
-- Transformers 4.16.2
-- Pytorch 1.10.2+cu102
-- Datasets 1.18.3
-- Tokenizers 0.11.0
+The use of this model needs to adhere to [this license from the Danish Parliament](https://www.ft.dk/da/aktuelt/tv-fra-folketinget/deling-og-rettigheder).