5-epochs5-char-based-freeze_cnn-dropout0.3

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: nan
• 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: 4.000000000000001e-06
• train_batch_size: 10
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 40
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
10.4529	0.07	2500	21.5863	1.0
5.5789	0.14	5000	7.7088	1.0
3.9797	0.2	7500	4.7691	1.0
3.5344	0.27	10000	3.7288	1.0
3.5794	0.34	12500	3.8970	1.0
3.5264	0.41	15000	3.9099	1.0
3.5346	0.47	17500	3.9135	1.0
3.6184	0.54	20000	3.9921	1.0
3.5259	0.61	22500	3.9837	1.0
3.5744	0.68	25000	3.9882	1.0
3.5945	0.74	27500	3.7869	1.0
3.6836	0.81	30000	3.9407	1.0
3.7874	0.88	32500	3.7358	1.0
3.7926	0.95	35000	3.7358	1.0
3.7937	1.01	37500	3.7358	1.0
3.7933	1.08	40000	3.7358	1.0
3.7923	1.15	42500	3.7358	1.0
3.7923	1.22	45000	3.7358	1.0
3.7965	1.28	47500	3.7358	1.0
3.7942	1.35	50000	3.7358	1.0
3.7934	1.42	52500	3.7358	1.0
3.7851	1.49	55000	3.7358	1.0
3.7826	1.55	57500	3.7358	1.0
3.7891	1.62	60000	3.7358	1.0
3.7992	1.69	62500	3.7358	1.0
3.8099	1.76	65000	3.7358	1.0
3.7788	1.82	67500	3.7358	1.0
3.7816	1.89	70000	3.7358	1.0
3.7846	1.96	72500	3.7358	1.0
3.783	2.03	75000	3.7358	1.0
3.8451	2.09	77500	3.7358	1.0
3.7883	2.16	80000	3.7358	1.0
3.7915	2.23	82500	3.7358	1.0
3.7688	2.3	85000	3.7358	1.0
3.7959	2.36	87500	3.7358	1.0
3.7794	2.43	90000	3.7358	1.0
3.7862	2.5	92500	3.7358	1.0
3.8008	2.57	95000	3.7358	1.0
3.793	2.63	97500	3.7358	1.0
3.7781	2.7	100000	3.7358	1.0
3.7878	2.77	102500	3.7358	1.0
3.836	2.84	105000	3.7358	1.0
3.7914	2.9	107500	3.7358	1.0
3.7886	2.97	110000	3.7358	1.0
3.7924	3.04	112500	3.7358	1.0
3.7866	3.11	115000	3.7358	1.0
3.8027	3.17	117500	3.7358	1.0
0.0	3.24	120000	nan	1.0
0.0	3.31	122500	nan	1.0
0.0	3.38	125000	nan	1.0
0.0	3.44	127500	nan	1.0
0.0	3.51	130000	nan	1.0
0.0	3.58	132500	nan	1.0
0.0	3.65	135000	nan	1.0
0.0	3.71	137500	nan	1.0
0.0	3.78	140000	nan	1.0
0.0	3.85	142500	nan	1.0
0.0	3.92	145000	nan	1.0
0.0	3.98	147500	nan	1.0
0.0	4.05	150000	nan	1.0
0.0	4.12	152500	nan	1.0
0.0	4.19	155000	nan	1.0
0.0	4.26	157500	nan	1.0
0.0	4.32	160000	nan	1.0
0.0	4.39	162500	nan	1.0
0.0	4.46	165000	nan	1.0
0.0	4.53	167500	nan	1.0
0.0	4.59	170000	nan	1.0
0.0	4.66	172500	nan	1.0
0.0	4.73	175000	nan	1.0
0.0	4.8	177500	nan	1.0
0.0	4.86	180000	nan	1.0
0.0	4.93	182500	nan	1.0
0.0	5.0	185000	nan	1.0

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1
• Datasets 2.14.5
• Tokenizers 0.14.1