tinybert_base_train_book_ent_15p_s_init_kd_complete_stsb

This model is a fine-tuned version of gokulsrinivasagan/tinybert_base_train_book_ent_15p_s_init_kd_complete on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 0.7455
• Pearson: 0.8187
• Spearmanr: 0.8154
• Combined Score: 0.8170

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: 256
• eval_batch_size: 256
• seed: 10
• 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
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Pearson	Spearmanr	Combined Score
2.6257	1.0	23	2.5060	0.1353	0.1309	0.1331
1.8313	2.0	46	1.6575	0.6385	0.6265	0.6325
1.1806	3.0	69	1.3400	0.7092	0.7236	0.7164
0.8641	4.0	92	1.0966	0.7756	0.7786	0.7771
0.7115	5.0	115	0.8398	0.7931	0.7898	0.7914
0.6248	6.0	138	0.7820	0.8130	0.8090	0.8110
0.5846	7.0	161	0.7455	0.8187	0.8154	0.8170
0.4653	8.0	184	0.8070	0.8201	0.8177	0.8189
0.4188	9.0	207	0.7894	0.8156	0.8131	0.8143
0.3692	10.0	230	0.8148	0.8154	0.8138	0.8146
0.3428	11.0	253	1.1896	0.8115	0.8174	0.8145
0.3529	12.0	276	0.9953	0.8173	0.8180	0.8176

Framework versions

• Transformers 4.51.2
• Pytorch 2.6.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.1