distilbert_lda_50_v1_stsb

This model is a fine-tuned version of gokulsrinivasagan/distilbert_lda_50_v1 on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 1.3807
• Pearson: 0.6576
• Spearmanr: 0.6483
• Combined Score: 0.6529

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
2.6455	1.0	23	2.4817	0.0933	0.0791	0.0862
2.0015	2.0	46	2.5968	0.1340	0.1191	0.1265
1.875	3.0	69	2.2990	0.2213	0.2184	0.2198
1.4576	4.0	92	2.3398	0.4446	0.4147	0.4296
1.0618	5.0	115	1.8954	0.5425	0.5250	0.5338
0.7001	6.0	138	1.9422	0.5693	0.5653	0.5673
0.5415	7.0	161	2.0586	0.5778	0.5692	0.5735
0.4005	8.0	184	1.5390	0.6166	0.6056	0.6111
0.3147	9.0	207	1.7006	0.5953	0.5793	0.5873
0.2829	10.0	230	1.4072	0.6480	0.6375	0.6428
0.2308	11.0	253	1.5357	0.6279	0.6134	0.6207
0.1996	12.0	276	1.6600	0.6341	0.6212	0.6276
0.1822	13.0	299	1.5625	0.6463	0.6354	0.6409
0.1646	14.0	322	1.3891	0.6635	0.6566	0.6601
0.1492	15.0	345	1.3807	0.6576	0.6483	0.6529
0.1296	16.0	368	1.4665	0.6426	0.6319	0.6372
0.1306	17.0	391	1.5261	0.6364	0.6213	0.6288
0.1205	18.0	414	1.4115	0.6427	0.6266	0.6347
0.1087	19.0	437	1.5183	0.6400	0.6248	0.6324
0.1072	20.0	460	1.5816	0.6414	0.6310	0.6362

Framework versions

• Transformers 4.46.3
• Pytorch 2.2.1+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3