wavlm_torgo_0H

This model is a fine-tuned version of microsoft/wavlm-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.2230
• 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: 0.0001
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
36.5759	0.1882	500	5.7798	1.0
4.1355	0.3764	1000	4.3661	1.0
3.9484	0.5645	1500	4.2577	1.0
3.6159	0.7527	2000	4.1272	1.0
3.6944	0.9409	2500	3.9745	1.0
3.8285	1.1291	3000	4.0134	1.0
3.6116	1.3173	3500	4.1692	1.0
3.5828	1.5055	4000	4.0013	1.0
3.5703	1.6936	4500	4.1055	1.0
3.5841	1.8818	5000	4.1041	1.0
3.8079	2.0700	5500	4.1574	1.0
3.5977	2.2582	6000	4.3217	1.0
3.5523	2.4464	6500	4.1800	1.0
3.5661	2.6346	7000	4.2053	1.0
3.5676	2.8227	7500	4.3885	1.0
3.794	3.0109	8000	4.2958	1.0
3.5647	3.1991	8500	4.2959	1.0
3.5805	3.3873	9000	4.3383	1.0
3.5475	3.5755	9500	4.1639	1.0
3.5523	3.7636	10000	4.2241	1.0
3.5982	3.9518	10500	4.3270	1.0
3.7088	4.1400	11000	4.2886	1.0
3.561	4.3282	11500	4.2801	1.0
3.5367	4.5164	12000	4.6914	1.0
3.5573	4.7046	12500	4.2071	1.0
3.5613	4.8927	13000	4.4513	1.0
3.719	5.0809	13500	4.3972	1.0
3.5376	5.2691	14000	4.3590	1.0
3.5313	5.4573	14500	4.3130	1.0
3.5384	5.6455	15000	4.4599	1.0
3.5755	5.8336	15500	4.3602	1.0
3.6912	6.0218	16000	4.2520	1.0
3.532	6.2100	16500	4.2731	1.0
3.565	6.3982	17000	4.2608	1.0
3.5328	6.5864	17500	4.2221	1.0
3.5361	6.7746	18000	4.2500	1.0
3.4975	6.9627	18500	4.2042	1.0
3.6749	7.1509	19000	4.2319	1.0
3.5316	7.3391	19500	4.2101	1.0
3.5262	7.5273	20000	4.2657	1.0
3.6605	7.7155	20500	4.2559	1.0
3.528	7.9037	21000	4.2230	1.0

Framework versions

• Transformers 4.40.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1