Update README.md

README.md

@@ -1,109 +1,27 @@
 ---
+language:
+- da
 license: cc-by-nc-4.0
-tags:
-- generated_from_trainer
 datasets:
-- common_voice
+- voxpopuli
+- common_voice_8_0
 model-index:
 - name: voxpopuli-wav2vec2-large-cv8-da
   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. -->
-
-# voxpopuli-wav2vec2-large-cv8-da
-
-This model is a fine-tuned version of [facebook/wav2vec2-large-sv-voxpopuli](https://huggingface.co/facebook/wav2vec2-large-sv-voxpopuli) on the common_voice dataset.
-It achieves the following results on the evaluation set:
-- Loss: 415.8608
-- Wer: 0.4709
+# VoxPopuli-Wav2vec2-CV8-da
 
 ## Model description
 
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
-
-## Training procedure
-
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- learning_rate: 4e-05
-- train_batch_size: 4
-- eval_batch_size: 8
-- seed: 4242
-- gradient_accumulation_steps: 8
-- 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: 500
-- num_epochs: 500
-- mixed_precision_training: Native AMP
-
-### Training results
+This model is a fine-tuned version of the Swedish acoustic model [facebook/wav2vec2-large-sv-voxpopuli](https://huggingface.co/facebook/wav2vec2-large-sv-voxpopuli) on the Danish part of [Common Voice 8.0](https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0), containing ~6 crowdsourced hours of read-aloud Danish speech.
 
-| Training Loss | Epoch  | Step  | Validation Loss | Wer    |
-|:-------------:|:------:|:-----:|:---------------:|:------:|
-| 726.4764      | 5.55   | 300   | 1339.8062       | 1.0    |
-| 574.6736      | 11.11  | 600   | 1265.3354       | 1.0    |
-| 534.5717      | 16.66  | 900   | 1108.4810       | 1.0    |
-| 343.2175      | 22.22  | 1200  | 669.4202        | 0.9374 |
-| 266.4228      | 27.77  | 1500  | 522.6546        | 0.8454 |
-| 220.443       | 33.33  | 1800  | 454.1028        | 0.7661 |
-| 185.4214      | 38.88  | 2100  | 413.6853        | 0.7089 |
-| 162.6397      | 44.44  | 2400  | 388.8118        | 0.6620 |
-| 142.0146      | 49.99  | 2700  | 368.1696        | 0.6286 |
-| 132.324       | 55.55  | 3000  | 368.3901        | 0.6071 |
-| 117.4966      | 61.11  | 3300  | 356.1192        | 0.5831 |
-| 106.5261      | 66.66  | 3600  | 349.3229        | 0.5655 |
-| 95.7354       | 72.22  | 3900  | 353.2257        | 0.5534 |
-| 91.4711       | 77.77  | 4200  | 351.9698        | 0.5453 |
-| 86.1983       | 83.33  | 4500  | 354.0186        | 0.5454 |
-| 77.3262       | 88.88  | 4800  | 355.5745        | 0.5390 |
-| 73.6696       | 94.44  | 5100  | 352.2862        | 0.5282 |
-| 69.1898       | 99.99  | 5400  | 351.4298        | 0.5223 |
-| 66.8595       | 105.55 | 5700  | 361.5981        | 0.5151 |
-| 61.6978       | 111.11 | 6000  | 362.9925        | 0.5097 |
-| 58.1949       | 116.66 | 6300  | 366.9365        | 0.5060 |
-| 54.6844       | 122.22 | 6600  | 361.5234        | 0.5074 |
-| 54.6381       | 127.77 | 6900  | 366.7358        | 0.5045 |
-| 52.0427       | 133.33 | 7200  | 373.6508        | 0.4990 |
-| 48.9871       | 138.88 | 7500  | 367.0853        | 0.4964 |
-| 46.3138       | 144.44 | 7800  | 375.1443        | 0.4979 |
-| 44.4403       | 149.99 | 8100  | 379.8337        | 0.4925 |
-| 45.3876       | 155.55 | 8400  | 382.2239        | 0.4862 |
-| 42.0502       | 161.11 | 8700  | 384.0788        | 0.4867 |
-| 40.0308       | 166.66 | 9000  | 383.6678        | 0.4802 |
-| 39.1281       | 172.22 | 9300  | 386.4395        | 0.4814 |
-| 37.5134       | 177.77 | 9600  | 390.5568        | 0.4752 |
-| 38.0693       | 183.33 | 9900  | 389.8726        | 0.4786 |
-| 35.0237       | 188.88 | 10200 | 391.2879        | 0.4725 |
-| 36.2227       | 194.44 | 10500 | 383.9153        | 0.4776 |
-| 34.0764       | 199.99 | 10800 | 406.2789        | 0.4798 |
-| 33.658        | 205.55 | 11100 | 407.3349        | 0.4721 |
-| 32.4348       | 211.11 | 11400 | 394.2111        | 0.4761 |
-| 30.4748       | 216.66 | 11700 | 406.4054        | 0.4720 |
-| 30.9136       | 222.22 | 12000 | 403.6569        | 0.4739 |
-| 31.7569       | 227.77 | 12300 | 406.9623        | 0.4702 |
-| 29.7905       | 233.33 | 12600 | 407.8719        | 0.4647 |
-| 28.9987       | 238.88 | 12900 | 419.9362        | 0.4690 |
-| 27.9413       | 244.44 | 13200 | 420.0189        | 0.4727 |
-| 27.4724       | 249.99 | 13500 | 408.5531        | 0.4676 |
-| 27.7707       | 255.55 | 13800 | 421.7307        | 0.4678 |
-| 27.2706       | 261.11 | 14100 | 415.8608        | 0.4709 |
 
+## Performance
 
-### Framework versions
+The model achieves the following WER scores (lower is better):
 
-- Transformers 4.16.2
-- Pytorch 1.10.2+cu102
-- Datasets 1.18.3
-- Tokenizers 0.11.0
+| **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) | 48.04 | xx.xx |
+| [Alvenir test set](https://huggingface.co/datasets/Alvenir/alvenir_asr_da_eval) | 48.43 | xx.xx |