82cbc615f68992009ef97934d22fc47e

This model is a fine-tuned version of studio-ousia/mluke-base-lite on the nyu-mll/glue [stsb] dataset. It achieves the following results on the evaluation set:

• Loss: 0.5573
• Data Size: 1.0
• Epoch Runtime: 18.5911
• Mse: 0.5575
• Mae: 0.5660
• R2: 0.7506

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	7.4972	0	1.7240	7.4984	2.3071	-2.3543
No log	1	179	4.2388	0.0078	2.1328	4.2399	1.7033	-0.8966
No log	2	358	2.3911	0.0156	2.2266	2.3921	1.3189	-0.0701
No log	3	537	2.5899	0.0312	2.7704	2.5904	1.3117	-0.1588
No log	4	716	1.1495	0.0625	3.4180	1.1499	0.8898	0.4856
No log	5	895	0.8822	0.125	4.5523	0.8826	0.7367	0.6052
0.1108	6	1074	0.8449	0.25	6.7020	0.8452	0.6917	0.6219
0.6718	7	1253	0.6184	0.5	10.8853	0.6187	0.6195	0.7232
0.5309	8.0	1432	0.5534	1.0	18.9865	0.5537	0.5678	0.7523
0.3487	9.0	1611	0.6084	1.0	18.9105	0.6086	0.5820	0.7278
0.3019	10.0	1790	0.5351	1.0	19.4072	0.5353	0.5542	0.7605
0.2017	11.0	1969	0.5369	1.0	20.0469	0.5371	0.5562	0.7597
0.1594	12.0	2148	0.5108	1.0	20.6199	0.5110	0.5570	0.7714
0.1406	13.0	2327	0.5344	1.0	20.6719	0.5345	0.5529	0.7609
0.1265	14.0	2506	0.5434	1.0	21.3759	0.5435	0.5568	0.7569
0.1089	15.0	2685	0.5361	1.0	18.3115	0.5364	0.5528	0.7600
0.0935	16.0	2864	0.5573	1.0	18.5911	0.5575	0.5660	0.7506

Framework versions

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