metadata

license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
  - generated_from_trainer
datasets:
  - common_voice_17_0
metrics:
  - wer
model-index:
  - name: xls-r-300m-hbs-fr-unfrozen-batch16
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_17_0
          type: common_voice_17_0
          config: hsb
          split: test
          args: hsb
        metrics:
          - name: Wer
            type: wer
            value: 0.3959700093720712

xls-r-300m-hbs-fr-unfrozen-batch16

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.7093
Wer: 0.3960
Cer: 0.0915

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: 0.0003
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
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: 100
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.5652	3.2258	100	3.3748	1.0	1.0
3.2583	6.4516	200	3.2149	1.0	1.0
3.1829	9.6774	300	3.1452	1.0	1.0
0.7256	12.9032	400	0.7889	0.7134	0.1766
0.3062	16.1290	500	0.6745	0.6146	0.1423
0.1843	19.3548	600	0.6301	0.5265	0.1242
0.1259	22.5806	700	0.6102	0.4820	0.1121
0.1386	25.8065	800	0.6702	0.4939	0.1176
0.0962	29.0323	900	0.6297	0.4806	0.1147
0.069	32.2581	1000	0.6766	0.4740	0.1113
0.0779	35.4839	1100	0.6565	0.4609	0.1075
0.0715	38.7097	1200	0.6649	0.4649	0.1103
0.0448	41.9355	1300	0.6558	0.4642	0.1094
0.0552	45.1613	1400	0.6893	0.4412	0.1035
0.0396	48.3871	1500	0.7179	0.4527	0.1041
0.0592	51.6129	1600	0.6455	0.4285	0.0976
0.0509	54.8387	1700	0.6605	0.4349	0.1005
0.0665	58.0645	1800	0.7340	0.4243	0.0991
0.0391	61.2903	1900	0.7378	0.4330	0.1018
0.0974	64.5161	2000	0.6984	0.4306	0.1003
0.0344	67.7419	2100	0.6895	0.4208	0.0974
0.043	70.9677	2200	0.7214	0.4140	0.0965
0.0248	74.1935	2300	0.7242	0.4149	0.0990
0.0194	77.4194	2400	0.7233	0.4107	0.0962
0.0277	80.6452	2500	0.7247	0.4100	0.0946
0.0447	83.8710	2600	0.7078	0.4004	0.0941
0.0291	87.0968	2700	0.7073	0.4002	0.0915
0.0208	90.3226	2800	0.7121	0.4025	0.0921
0.0278	93.5484	2900	0.6998	0.3932	0.0914
0.0569	96.7742	3000	0.7105	0.3964	0.0918
0.0132	100.0	3100	0.7093	0.3960	0.0915

Framework versions

Transformers 4.42.0.dev0
Pytorch 2.3.1+cu121
Datasets 2.19.2
Tokenizers 0.19.1