w2v-bert-2.0-lg-CV-Fleurs-filtered-100hrs-v11

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.4163
• Wer: 0.4201
• Cer: 0.0816

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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_ratio: 0.1
• num_epochs: 70
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.8755	1.0	7125	0.3901	0.4546	0.0912
0.1904	2.0	14250	0.3507	0.4646	0.0896
0.1641	3.0	21375	0.3312	0.4389	0.0854
0.1511	4.0	28500	0.3229	0.4311	0.0802
0.145	5.0	35625	0.3518	0.4273	0.0820
0.1406	6.0	42750	0.3275	0.4302	0.0812
0.1379	7.0	49875	0.3627	0.5317	0.0944
0.1316	8.0	57000	0.3234	0.4175	0.0817
0.1182	9.0	64125	0.3534	0.4317	0.0819
0.1082	10.0	71250	0.3230	0.4053	0.0777
0.0976	11.0	78375	0.3261	0.4194	0.0791
0.0884	12.0	85500	0.3183	0.4119	0.0798
0.0803	13.0	92625	0.3695	0.4170	0.0791
0.072	14.0	99750	0.3596	0.4102	0.0799
0.0637	15.0	106875	0.3625	0.4137	0.0803
0.0554	16.0	114000	0.3958	0.4336	0.0822
0.0481	17.0	121125	0.3820	0.4128	0.0796
0.0422	18.0	128250	0.4239	0.4134	0.0807
0.0357	19.0	135375	0.3978	0.4436	0.0839
0.031	20.0	142500	0.4163	0.4201	0.0816

Framework versions

• Transformers 4.46.2
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.3