MSP-AVSR

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

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 500
training_steps: 10000

Training Loss	Epoch	Step	Validation Loss	Wer
2.3649	0.05	500	1.1951	0.2032
2.3195	0.1	1000	1.3344	0.2257
2.2183	0.15	1500	1.4732	0.2762
2.3529	0.2	2000	1.3084	0.2243
2.1855	0.25	2500	1.4419	0.2512
2.1410	0.3	3000	1.2315	0.2163
1.9520	0.35	3500	1.2516	0.2175
2.1277	0.4	4000	1.2095	0.2108
2.1164	0.45	4500	1.2529	0.2251
2.3793	0.5	5000	1.2943	0.2208
2.0423	0.55	5500	1.1921	0.1954
2.1330	0.6	6000	1.2455	0.2142
2.0254	0.65	6500	1.1977	0.1988
2.0638	0.7	7000	1.3258	0.2297
2.0226	0.75	7500	1.3211	0.2253
2.2216	0.8	8000	1.4072	0.2438
1.9457	0.85	8500	1.2661	0.2148
2.0084	0.9	9000	1.3744	0.2348
2.0407	0.95	9500	1.3376	0.2270
1.9916	1.0	10000	1.3328	0.2258