dc5ddebf38b1edec0d7141a6be8c9449

This model is a fine-tuned version of facebook/opt-125m on the nyu-mll/glue [stsb] dataset. It achieves the following results on the evaluation set:

• Loss: 0.5430
• Data Size: 1.0
• Epoch Runtime: 14.4102
• Mse: 0.5432
• Mae: 0.5796
• R2: 0.7570

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	12.0061	0	2.0216	12.0075	3.1072	-4.3714
No log	1	179	4.2542	0.0078	2.1488	4.2548	1.6555	-0.9033
No log	2	358	2.1209	0.0156	2.1017	2.1216	1.2195	0.0509
No log	3	537	1.7771	0.0312	2.4676	1.7776	1.0859	0.2048
No log	4	716	1.5141	0.0625	2.9663	1.5142	0.9685	0.3226
No log	5	895	0.9416	0.125	3.7103	0.9418	0.7702	0.5787
0.0902	6	1074	0.6281	0.25	5.1667	0.6283	0.6200	0.7190
0.585	7	1253	0.6868	0.5	8.1877	0.6872	0.6646	0.6926
0.4427	8.0	1432	0.5980	1.0	14.2426	0.5982	0.6123	0.7324
0.2646	9.0	1611	0.6560	1.0	14.1542	0.6563	0.6269	0.7064
0.1816	10.0	1790	0.5479	1.0	14.2617	0.5481	0.5846	0.7548
0.1529	11.0	1969	0.5540	1.0	14.4668	0.5540	0.5794	0.7522
0.1128	12.0	2148	0.6138	1.0	14.6348	0.6139	0.6183	0.7254
0.1129	13.0	2327	0.5106	1.0	14.7767	0.5108	0.5556	0.7715
0.0825	14.0	2506	0.5853	1.0	14.7161	0.5854	0.6061	0.7381
0.0761	15.0	2685	0.5200	1.0	14.8478	0.5201	0.5579	0.7673
0.0677	16.0	2864	0.6197	1.0	14.3931	0.6198	0.6250	0.7227
0.0608	17.0	3043	0.5430	1.0	14.4102	0.5432	0.5796	0.7570

Framework versions

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