README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: working
    results: []

working

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2536
• Wer: 0.5805
• Cer: 0.0690

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: 3e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• training_steps: 26000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.0464	1.49	1000	1.5865	0.9968	0.5743
1.2779	2.98	2000	0.4195	0.7539	0.1008
1.0865	4.46	3000	0.3272	0.6791	0.0852
1.0852	5.95	4000	0.3039	0.6409	0.0789
1.0193	7.44	5000	0.3009	0.6442	0.0809
0.9566	8.93	6000	0.2804	0.6182	0.0762
0.9086	10.42	7000	0.2842	0.6336	0.0772
0.9161	11.9	8000	0.2757	0.6044	0.0735
0.8569	13.39	9000	0.2809	0.6052	0.0748
0.8517	14.88	10000	0.2813	0.6166	0.0759
0.8304	16.37	11000	0.2808	0.6044	0.0759
0.8325	17.86	12000	0.2656	0.6060	0.0735
0.7918	19.35	13000	0.2676	0.5930	0.0730
0.8061	20.83	14000	0.2740	0.5890	0.0731
0.7206	22.32	15000	0.2735	0.6141	0.0730
0.7372	23.81	16000	0.2663	0.5857	0.0711
0.7285	25.3	17000	0.2708	0.5841	0.0723
0.6778	26.79	18000	0.2726	0.5825	0.0724
0.7117	28.27	19000	0.2737	0.5898	0.0723
0.6951	29.76	20000	0.2733	0.5890	0.0721
0.6775	31.25	21000	0.2721	0.5865	0.0725
0.666	32.74	22000	0.2733	0.5857	0.0721
0.688	34.23	23000	0.2700	0.5890	0.0718
0.6813	35.71	24000	0.2724	0.5930	0.0717
0.6409	37.2	25000	0.2704	0.5922	0.0714
0.6427	38.69	26000	0.2704	0.5881	0.0713

Framework versions

• Transformers 4.17.0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0