YAML Metadata Error: "language[0]" must only contain lowercase characters
YAML Metadata Error: "language[0]" with value "en-US" is not valid. It must be an ISO 639-1, 639-2 or 639-3 code (two/three letters), or a special value like "code", "multilingual". If you want to use BCP-47 identifiers, you can specify them in language_bcp47.

xtreme_s_xlsr_300m_minds14.en-US_2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the GOOGLE/XTREME_S - MINDS14.EN-US dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5685
  • F1: 0.8747
  • Accuracy: 0.8759

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: 2
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 32
  • total_eval_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 50.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss F1 Accuracy
2.6195 3.95 20 2.6348 0.0172 0.0816
2.5925 7.95 40 2.6119 0.0352 0.0851
2.1271 11.95 60 2.3066 0.1556 0.1986
1.2618 15.95 80 1.3810 0.6877 0.7128
0.5455 19.95 100 1.0403 0.6992 0.7270
0.2571 23.95 120 0.8423 0.8160 0.8121
0.3478 27.95 140 0.6500 0.8516 0.8440
0.0732 31.95 160 0.7066 0.8123 0.8156
0.1092 35.95 180 0.5878 0.8767 0.8759
0.0271 39.95 200 0.5994 0.8578 0.8617
0.4664 43.95 220 0.7830 0.8403 0.8440
0.0192 47.95 240 0.5685 0.8747 0.8759

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.11.0+cu113
  • Datasets 2.1.0
  • Tokenizers 0.12.1
Downloads last month
15
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.