metadata

license: cc-by-nc-4.0
base_model: facebook/mms-1b-all
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: mms-MGB3
    results: []

mms-MGB3

This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.9382
Wer: 0.6591

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: 1e-05
train_batch_size: 14
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_steps: 50
num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
9.7535	0.13	250	8.6735	1.0023
3.2385	0.27	500	3.3341	1.0003
2.3512	0.4	750	2.0937	0.9027
1.4967	0.53	1000	1.3694	0.7637
1.3214	0.67	1250	1.2237	0.7347
1.2072	0.8	1500	1.1672	0.7176
1.1913	0.93	1750	1.1334	0.7108
1.1127	1.07	2000	1.1102	0.7044
1.1454	1.2	2250	1.0919	0.6996
1.1128	1.33	2500	1.0763	0.6955
1.086	1.47	2750	1.0629	0.6916
1.1285	1.6	3000	1.0503	0.6888
1.081	1.73	3250	1.0406	0.6886
1.0449	1.86	3500	1.0320	0.6857
1.0625	2.0	3750	1.0231	0.6849
1.0892	2.13	4000	1.0157	0.6824
1.0566	2.26	4250	1.0097	0.6795
1.0972	2.4	4500	1.0036	0.6747
1.0617	2.53	4750	0.9957	0.6744
1.0441	2.66	5000	0.9881	0.6756
1.0589	2.8	5250	0.9807	0.6718
1.0005	2.93	5500	0.9758	0.6713
1.0447	3.06	5750	0.9701	0.6694
0.9722	3.2	6000	0.9667	0.6664
0.9873	3.33	6250	0.9595	0.6675
0.9857	3.46	6500	0.9551	0.6633
0.9625	3.6	6750	0.9519	0.6633
0.9748	3.73	7000	0.9464	0.6607
0.9626	3.86	7250	0.9427	0.6617
1.0242	4.0	7500	0.9382	0.6591

Framework versions

Transformers 4.33.2
Pytorch 2.0.1
Datasets 2.19.1
Tokenizers 0.13.3