Wave2Vec2_OV_Vie

This model is a fine-tuned version of facebook/wav2vec2-base on the VIVOS - NA dataset. It achieves the following results on the evaluation set:

• Loss: 3.5894
• Wer: 1.0

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.27	100	3.9210	1.0
No log	0.55	200	3.4375	1.0
No log	0.82	300	3.4356	1.0
No log	1.1	400	3.4045	1.0
4.1866	1.37	500	3.4694	1.0
4.1866	1.65	600	3.6266	1.0
4.1866	1.92	700	3.5694	1.0
4.1866	2.19	800	3.5733	1.0
4.1866	2.47	900	3.6381	1.0
3.4376	2.74	1000	3.6604	1.0
3.4376	3.02	1100	3.5868	1.0
3.4376	3.29	1200	3.4988	1.0
3.4376	3.57	1300	3.5409	1.0
3.4376	3.84	1400	3.4883	1.0
3.4365	4.12	1500	3.6125	1.0
3.4365	4.39	1600	3.6123	1.0
3.4365	4.66	1700	3.5978	1.0
3.4365	4.94	1800	3.5693	1.0
3.4365	5.21	1900	3.5659	1.0
3.4339	5.49	2000	3.6234	1.0
3.4339	5.76	2100	3.5997	1.0
3.4339	6.04	2200	3.6529	1.0
3.4339	6.31	2300	3.5780	1.0
3.4339	6.58	2400	3.5844	1.0
3.4333	6.86	2500	3.5792	1.0
3.4333	7.13	2600	3.5468	1.0
3.4333	7.41	2700	3.5691	1.0
3.4333	7.68	2800	3.5408	1.0
3.4333	7.96	2900	3.5482	1.0
3.4294	8.23	3000	3.6070	1.0
3.4294	8.5	3100	3.5905	1.0
3.4294	8.78	3200	3.6018	1.0
3.4294	9.05	3300	3.6326	1.0
3.4294	9.33	3400	3.6214	1.0
3.4293	9.6	3500	3.6372	1.0
3.4293	9.88	3600	3.6215	1.0
3.4293	10.15	3700	3.5106	1.0
3.4293	10.43	3800	3.5066	1.0
3.4293	10.7	3900	3.5352	1.0
3.4295	10.97	4000	3.5129	1.0
3.4295	11.25	4100	3.6384	1.0
3.4295	11.52	4200	3.6019	1.0
3.4295	11.8	4300	3.5876	1.0
3.4295	12.07	4400	3.6207	1.0
3.4252	12.35	4500	3.5998	1.0
3.4252	12.62	4600	3.6216	1.0
3.4252	12.89	4700	3.6073	1.0
3.4252	13.17	4800	3.5567	1.0
3.4252	13.44	4900	3.5745	1.0
3.4274	13.72	5000	3.5738	1.0
3.4274	13.99	5100	3.5914	1.0
3.4274	14.27	5200	3.6004	1.0
3.4274	14.54	5300	3.5968	1.0
3.4274	14.81	5400	3.5908	1.0

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.0
• Tokenizers 0.13.3