Modelpublic

Mardiyyah/CeLLaTe-tapt-bioformer16l-tokenizer-original-spanmask-combinedData

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes23downloads

Model Card

CeLLaTe-tapt-bioformer16l-tokenizer-original-spanmask-combinedData

This model is a fine-tuned version of bioformers/bioformer-16l on the Mardiyyah/CombinedTAPTdataset dataset. It achieves the following results on the evaluation set:

—Loss: 2.4313
—Accuracy: 0.5592
—Perplexity: 11.3734

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: 3e-05
—trainbatchsize: 32
—evalbatchsize: 32
—seed: 3407
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 64
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-06 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_ratio: 0.06
—num_epochs: 100
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Perplexity
3.2395	1.0	29	2.7241	0.5309	15.2428
2.9615	2.0	58	2.6084	0.5384	13.5772
2.8711	3.0	87	2.5849	0.5398	13.2621
2.8244	4.0	116	2.5672	0.5422	13.0289
2.8053	5.0	145	2.5546	0.5471	12.8658
2.7995	6.0	174	2.5014	0.5477	12.2000
2.7511	7.0	203	2.5206	0.5434	12.4366
2.7662	8.0	232	2.4845	0.5516	11.9951
2.7266	9.0	261	2.5541	0.5451	12.8597
2.7294	10.0	290	2.4709	0.5538	11.8326
2.7071	11.0	319	2.5285	0.5480	12.5352
2.6804	12.0	348	2.4847	0.5493	11.9975
2.6619	13.0	377	2.5265	0.5459	12.5100
2.6361	14.0	406	2.4463	0.5548	11.5458
2.6503	15.0	435	2.4666	0.5548	11.7824
2.6363	16.0	464	2.4632	0.5513	11.7418
2.6221	17.0	493	2.4676	0.5567	11.7935
2.6248	18.0	522	2.4870	0.5486	12.0255
2.6232	19.0	551	2.4810	0.5500	11.9531

Framework versions

—Transformers 4.48.2
—Pytorch 2.4.1+cu121
—Datasets 3.0.2
—Tokenizers 0.21.0