learning_rate: 0.0004
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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: 132
num_epochs: 100
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
4.473	2.7027	200	2.4188	1.0
1.3356	5.4054	400	0.2886	0.5197
0.1766	8.1081	600	0.3087	0.3930
0.0958	10.8108	800	0.2624	0.3323
0.0596	13.5135	1000	0.3792	0.3440
0.0461	16.2162	1200	0.3037	0.3365
0.0352	18.9189	1400	0.3387	0.3387
0.0227	21.6216	1600	0.3182	0.3387
0.0243	24.3243	1800	0.3493	0.3450
0.0224	27.0270	2000	0.3503	0.3312
0.0145	29.7297	2200	0.3551	0.3301
0.0205	32.4324	2400	0.3310	0.3514
0.0126	35.1351	2600	0.3741	0.3323
0.0137	37.8378	2800	0.3761	0.3323
0.014	40.5405	3000	0.3646	0.3355
0.0096	43.2432	3200	0.3660	0.3301
0.0099	45.9459	3400	0.3745	0.3312
0.0121	48.6486	3600	0.3930	0.3291
0.0131	51.3514	3800	0.3886	0.3280
0.0095	54.0541	4000	0.3874	0.3450
0.0066	56.7568	4200	0.3877	0.3269
0.0047	59.4595	4400	0.3871	0.3248
0.0074	62.1622	4600	0.3886	0.3259
0.0068	64.8649	4800	0.4206	0.3259
0.005	67.5676	5000	0.4182	0.3227
0.0087	70.2703	5200	0.4100	0.3248
0.0047	72.9730	5400	0.4196	0.3259
0.0056	75.6757	5600	0.4133	0.3259
0.0048	78.3784	5800	0.4135	0.3269
0.0036	81.0811	6000	0.3901	0.3248
0.003	83.7838	6200	0.3869	0.3227
0.0021	86.4865	6400	0.3896	0.3227
0.0013	89.1892	6600	0.3893	0.3216
0.0019	91.8919	6800	0.3983	0.3216
0.0015	94.5946	7000	0.4023	0.3227
0.0013	97.2973	7200	0.3973	0.3227
0.0017	100.0	7400	0.3950	0.3216

susmitabhatt/xlsr-a-nose

xlsr-a-nose