breeze-listen-w2v2-id

This model is a fine-tuned version of facebook/mms-1b-all on the MOZILLA-FOUNDATION/COMMON_VOICE_16_0 - ID dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1253
  • Wer: 0.1458

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: 0.001
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: multi-GPU
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 4.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
No log 0.1 200 3.2671 1.0
No log 0.19 400 2.8741 1.0007
3.8381 0.29 600 2.7612 0.9955
3.8381 0.38 800 2.6333 0.9981
2.6996 0.48 1000 2.3074 0.9771
2.6996 0.58 1200 2.0155 0.9286
2.6996 0.67 1400 1.9155 0.8947
2.2919 0.77 1600 1.6412 0.8814
2.2919 0.87 1800 1.4531 0.8285
1.5872 0.96 2000 0.1813 0.2060
1.5872 1.06 2200 0.1636 0.1806
1.5872 1.15 2400 0.1558 0.1744
0.2659 1.25 2600 0.1522 0.1647
0.2659 1.35 2800 0.1553 0.1664
0.2436 1.44 3000 0.1841 0.1961
0.2436 1.54 3200 0.1419 0.1640
0.2436 1.64 3400 0.1456 0.1714
0.2464 1.73 3600 0.1402 0.1607
0.2464 1.83 3800 0.1345 0.1528
0.2292 1.92 4000 0.1342 0.1556
0.2292 2.02 4200 0.1334 0.1552
0.2292 2.12 4400 0.1352 0.1543
0.2209 2.21 4600 0.1350 0.1538
0.2209 2.31 4800 0.1342 0.1530
0.2136 2.41 5000 0.1320 0.1540
0.2136 2.5 5200 0.1369 0.1569
0.2136 2.6 5400 0.1314 0.1517
0.2154 2.69 5600 0.1304 0.1506
0.2154 2.79 5800 0.1320 0.1507
0.2123 2.89 6000 0.1319 0.1524
0.2123 2.98 6200 0.1292 0.1524
0.2123 3.08 6400 0.1283 0.1488
0.2109 3.18 6600 0.1258 0.1492
0.2109 3.27 6800 0.1291 0.1488
0.2103 3.37 7000 0.1278 0.1484
0.2103 3.46 7200 0.1250 0.1478
0.2103 3.56 7400 0.1277 0.1482
0.1986 3.66 7600 0.1256 0.1476
0.1986 3.75 7800 0.1258 0.1468
0.1954 3.85 8000 0.1256 0.1465
0.1954 3.95 8200 0.1253 0.1456

Framework versions

  • Transformers 4.38.0.dev0
  • Pytorch 2.1.2+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.1
Downloads last month
10
Safetensors
Model size
965M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for simpragma/breeze-listen-w2v2-id

Finetuned
(254)
this model

Evaluation results