Hubert-common_voice-ja-demo

This model is a fine-tuned version of rinna/japanese-hubert-base on the MOZILLA-FOUNDATION/COMMON_VOICE_13_0 - JA dataset. It achieves the following results on the evaluation set:

• Loss: 5.6933
• 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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.2660	100	54.1581	1.0
No log	0.5319	200	24.3299	1.0
No log	0.7979	300	5.7205	1.0
No log	1.0638	400	5.7282	1.0
23.1666	1.3298	500	5.7253	1.0
23.1666	1.5957	600	5.7277	1.0
23.1666	1.8617	700	5.6616	1.0
23.1666	2.1277	800	5.6682	1.0
23.1666	2.3936	900	5.6722	1.0
5.7461	2.6596	1000	5.6670	1.0
5.7461	2.9255	1100	5.6700	1.0
5.7461	3.1915	1200	5.6794	1.0
5.7461	3.4574	1300	5.7088	1.0
5.7461	3.7234	1400	5.7975	1.0
5.7384	3.9894	1500	5.6856	1.0
5.7384	4.2553	1600	5.6743	1.0
5.7384	4.5213	1700	5.6769	1.0
5.7384	4.7872	1800	5.6739	1.0
5.7384	5.0532	1900	5.7652	1.0
5.7316	5.3191	2000	5.7474	1.0
5.7316	5.5851	2100	5.7853	1.0
5.7316	5.8511	2200	5.7743	1.0
5.7316	6.1170	2300	5.6709	1.0
5.7316	6.3830	2400	5.6688	1.0
5.7285	6.6489	2500	5.6701	1.0
5.7285	6.9149	2600	5.6694	1.0
5.7285	7.1809	2700	5.7340	1.0
5.7285	7.4468	2800	5.7335	1.0
5.7285	7.7128	2900	5.7401	1.0
5.7214	7.9787	3000	5.6974	1.0
5.7214	8.2447	3100	5.6702	1.0
5.7214	8.5106	3200	5.6725	1.0
5.7214	8.7766	3300	5.6709	1.0
5.7214	9.0426	3400	5.7029	1.0
5.8418	9.3085	3500	5.6817	1.0
5.8418	9.5745	3600	5.6595	1.0
5.8418	9.8404	3700	5.6728	1.0
5.8418	10.1064	3800	5.6784	1.0
5.8418	10.3723	3900	5.6867	1.0
5.722	10.6383	4000	5.6920	1.0
5.722	10.9043	4100	5.6865	1.0
5.722	11.1702	4200	5.6617	1.0
5.722	11.4362	4300	5.6692	1.0
5.722	11.7021	4400	5.6652	1.0
5.7175	11.9681	4500	5.6653	1.0
5.7175	12.2340	4600	5.7036	1.0
5.7175	12.5	4700	5.7013	1.0
5.7175	12.7660	4800	5.7120	1.0
5.7175	13.0319	4900	5.6736	1.0
5.7121	13.2979	5000	5.6713	1.0
5.7121	13.5638	5100	5.6844	1.0
5.7121	13.8298	5200	5.6856	1.0
5.7121	14.0957	5300	5.7027	1.0
5.7121	14.3617	5400	5.6934	1.0
5.7053	14.6277	5500	5.6952	1.0
5.7053	14.8936	5600	5.6950	1.0

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3