wav2vec2-E30_freq_speed_pause2

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: 1.4385
• Cer: 30.1116

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
31.4976	0.1289	200	5.0615	100.0
4.9534	0.2579	400	4.6738	100.0
4.8189	0.3868	600	4.6468	100.0
4.7696	0.5158	800	4.5988	100.0
4.6915	0.6447	1000	4.5676	100.0
4.66	0.7737	1200	4.5294	97.9260
4.6295	0.9026	1400	4.5050	97.9201
4.5468	1.0316	1600	4.4362	97.9377
4.4058	1.1605	1800	4.3229	96.0928
4.0025	1.2895	2000	3.8060	82.1093
3.3806	1.4184	2200	3.2988	62.4442
2.9142	1.5474	2400	2.9776	55.4054
2.5748	1.6763	2600	2.4457	47.4207
2.3231	1.8053	2800	2.2643	44.4360
2.1253	1.9342	3000	2.0188	39.4536
1.9269	2.0632	3200	1.8784	37.6439
1.7746	2.1921	3400	1.7279	35.2879
1.6518	2.3211	3600	1.6686	33.9953
1.5782	2.4500	3800	1.6163	32.7203
1.5226	2.5790	4000	1.5312	31.2162
1.472	2.7079	4200	1.4718	30.2585
1.3968	2.8369	4400	1.4688	30.5934
1.3675	2.9658	4600	1.4385	30.1116

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3