hubert-base-ls960-finetuned-ic-slurp-wt_init-frz

This model is a fine-tuned version of facebook/hubert-base-ls960 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0889
• Accuracy: 0.4598

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: 24
• eval_batch_size: 24
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.6605	1.0	527	3.6385	0.1020
3.6135	2.0	1055	3.5710	0.1200
3.4222	3.0	1582	3.3394	0.1738
3.1948	4.0	2110	3.2132	0.2052
2.8791	5.0	2637	2.9508	0.2581
2.7807	6.0	3165	2.7201	0.3109
2.4647	7.0	3692	2.6056	0.3393
2.3009	8.0	4220	2.4893	0.3816
2.0953	9.0	4747	2.4874	0.3902
1.8074	10.0	5275	2.4705	0.4035
1.8209	11.0	5802	2.4465	0.4177
1.4822	12.0	6330	2.5310	0.4228
1.426	13.0	6857	2.5097	0.4305
1.2877	14.0	7385	2.5365	0.4368
1.0833	15.0	7912	2.5874	0.4404
1.0709	16.0	8440	2.6478	0.4373
0.8176	17.0	8967	2.7096	0.4409
0.803	18.0	9495	2.7965	0.4491
0.6678	19.0	10022	2.9335	0.4470
0.7066	20.0	10550	3.0013	0.4408
0.5935	21.0	11077	2.9613	0.4544
0.5703	22.0	11605	2.9915	0.4534
0.5	23.0	12132	3.0625	0.4556
0.55	24.0	12660	3.0889	0.4598
0.3977	25.0	13187	3.1962	0.4551
0.4578	26.0	13715	3.2863	0.4574
0.3343	27.0	14242	3.3401	0.4531
0.4414	28.0	14770	3.3229	0.4557
0.2551	29.0	15297	3.4294	0.4567

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2