You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

wav2vec2-large-mms-1b-mos-V2

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

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 0.001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 5
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
36.2909	0.1552	100	1.0946	0.9531
4.3897	0.3103	200	0.7441	0.7478
3.8096	0.4655	300	0.5952	0.5489
3.2801	0.6206	400	0.5921	0.6137
3.241	0.7758	500	0.5730	0.6903
3.2619	0.9310	600	0.5520	0.7171
3.0675	1.0861	700	0.5009	0.5014
2.8962	1.2413	800	0.4948	0.4978
2.9737	1.3964	900	0.4830	0.5002
2.871	1.5516	1000	0.4692	0.4823
2.8204	1.7067	1100	0.4531	0.4705
2.8753	1.8619	1200	0.4439	0.4750
2.7596	2.0171	1300	0.4533	0.4951
2.6663	2.1722	1400	0.4556	0.4711
2.5645	2.3274	1500	0.4173	0.4708
2.691	2.4825	1600	0.4225	0.4596
2.5496	2.6377	1700	0.4254	0.4634
2.5257	2.7929	1800	0.4058	0.4479
2.6506	2.9480	1900	0.4065	0.4623
2.5693	3.1032	2000	0.4015	0.4560
2.3695	3.2583	2100	0.3984	0.4462
2.4252	3.4135	2200	0.3942	0.4496
2.3974	3.5687	2300	0.3841	0.4477
2.344	3.7238	2400	0.3789	0.4385
2.4957	3.8790	2500	0.3721	0.4392
2.3832	4.0341	2600	0.3693	0.4376
2.2369	4.1893	2700	0.3684	0.4357
2.2326	4.3445	2800	0.3666	0.4368
2.3816	4.4996	2900	0.3634	0.4379
2.2641	4.6548	3000	0.3619	0.4345
2.279	4.8099	3100	0.3595	0.4329
2.179	4.9651	3200	0.3582	0.4330

Safetensors

Model size

1.0B params

Tensor type

F32

Base model

Finetuned

(382)

this model