README.md

metadata

license: mit
base_model: arslanarjumand/wav2vec-reptiles
tags:
  - generated_from_trainer
model-index:
  - name: wav2vec-repeat
    results: []

wav2vec-repeat

This model is a fine-tuned version of arslanarjumand/wav2vec-reptiles on the None dataset. It achieves the following results on the evaluation set:

• Loss: 205.9549
• Pcc Accuracy: 0.8004
• Pcc Fluency: 0.7759
• Pcc Total Score: 0.8207
• Pcc Content: 0.7220

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: 2.5e-05
• train_batch_size: 4
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.5
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Pcc Accuracy	Pcc Fluency	Pcc Total Score	Pcc Content
507.295	3.54	500	538.7184	0.2592	0.2368	0.2807	0.3206
267.4833	7.08	1000	374.0983	0.5787	0.5582	0.5900	0.5040
246.7156	10.62	1500	483.3237	0.6618	0.6387	0.6761	0.5837
269.7238	14.16	2000	446.4642	0.6964	0.6691	0.7131	0.6288
289.3261	17.7	2500	244.4726	0.7201	0.6928	0.7371	0.6482
249.89	21.24	3000	413.8036	0.7340	0.7052	0.7548	0.6796
235.8593	24.78	3500	251.3629	0.7472	0.7217	0.7676	0.6808
217.7143	28.32	4000	212.4162	0.7779	0.7547	0.7973	0.6948
123.7326	31.86	4500	362.4697	0.7782	0.7528	0.7987	0.7062
132.7905	35.4	5000	228.9714	0.7826	0.7603	0.8021	0.6987
111.7989	38.94	5500	189.2367	0.7985	0.7754	0.8188	0.7169
104.5979	42.48	6000	271.8181	0.7929	0.7692	0.8143	0.7192
115.256	46.02	6500	220.4324	0.8008	0.7753	0.8209	0.7230
86.3804	49.56	7000	205.9549	0.8004	0.7759	0.8207	0.7220

Framework versions

• Transformers 4.38.1
• Pytorch 2.1.2
• Datasets 2.17.1
• Tokenizers 0.15.2