66042c5d15c72970aacdd01c08465813

This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the nyu-mll/glue [stsb] dataset. It achieves the following results on the evaluation set:

• Loss: 0.6360
• Data Size: 1.0
• Epoch Runtime: 10.9811
• Mse: 0.6361
• Mae: 0.6071
• R2: 0.7154

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_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: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Mse	Mae	R2
No log	0	0	7.2959	0	1.2205	7.2971	2.2707	-2.2643
No log	1	179	3.8976	0.0078	1.5296	3.8987	1.6467	-0.7440
No log	2	358	2.2853	0.0156	1.6104	2.2860	1.2819	-0.0226
No log	3	537	2.5867	0.0312	2.0079	2.5877	1.3672	-0.1576
No log	4	716	2.1014	0.0625	2.3998	2.1022	1.2436	0.0596
No log	5	895	0.8921	0.125	2.9282	0.8926	0.7631	0.6007
0.1315	6	1074	0.7837	0.25	4.1671	0.7842	0.6791	0.6492
0.7823	7	1253	0.7277	0.5	6.5455	0.7283	0.6781	0.6742
0.6355	8.0	1432	0.6166	1.0	11.2578	0.6170	0.6052	0.7240
0.3982	9.0	1611	0.6060	1.0	11.2043	0.6064	0.5842	0.7287
0.2705	10.0	1790	0.6099	1.0	11.5508	0.6101	0.6025	0.7271
0.2075	11.0	1969	0.5823	1.0	11.2443	0.5828	0.5789	0.7393
0.1817	12.0	2148	0.6110	1.0	10.8460	0.6114	0.5984	0.7265
0.1452	13.0	2327	0.5857	1.0	10.9459	0.5860	0.5731	0.7379
0.128	14.0	2506	0.5893	1.0	10.9863	0.5895	0.5831	0.7363
0.1102	15.0	2685	0.6360	1.0	10.9811	0.6361	0.6071	0.7154

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1