wav2vec-reptiles

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: 182.3516
• Pcc Accuracy: 0.6684
• Pcc Fluency: 0.6499
• Pcc Total Score: 0.7110
• Pcc Content: 0.6788

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: 5.5e-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: 0.4
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Pcc Accuracy	Pcc Fluency	Pcc Total Score	Pcc Content
2719.4074	0.97	500	2790.7349	0.1171	0.1116	0.1218	0.1245
386.8535	1.93	1000	361.3293	0.1481	0.1332	0.1511	0.1445
273.8093	2.9	1500	304.4040	0.2869	0.2915	0.3062	0.2849
280.8214	3.87	2000	277.9273	0.4065	0.4344	0.4465	0.4131
264.1531	4.84	2500	265.5385	0.5012	0.5234	0.5490	0.5117
211.6362	5.8	3000	226.9335	0.5675	0.5768	0.6171	0.5817
217.8737	6.77	3500	218.1019	0.6089	0.5984	0.6525	0.6194
180.3319	7.74	4000	201.4108	0.6296	0.6142	0.6721	0.6395
174.7695	8.7	4500	201.3474	0.6427	0.6297	0.6872	0.6542
182.4466	9.67	5000	189.6567	0.6566	0.6333	0.6957	0.6619
184.7177	10.64	5500	182.7654	0.6628	0.6405	0.7033	0.6713
174.6915	11.61	6000	181.2284	0.6635	0.6479	0.7077	0.6755
187.671	12.57	6500	180.5753	0.6676	0.6486	0.7099	0.6773
166.4409	13.54	7000	181.2506	0.6682	0.6493	0.7105	0.6781
176.7043	14.51	7500	182.3516	0.6684	0.6499	0.7110	0.6788

Framework versions

• Transformers 4.37.0
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.1