CeLLaTe-tapt-bioformer16l-tokenizer-adapted-spanmask

This model is a fine-tuned version of bioformers/bioformer-16l on the Mardiyyah/TAPT_CeLLaTe2.0 dataset. It achieves the following results on the evaluation set:

• Loss: 2.7320
• Accuracy: 0.5272
• Perplexity: 15.3632

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
• train_batch_size: 32
• eval_batch_size: 32
• seed: 3407
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Perplexity
3.6895	1.0	15	3.1710	0.5083	23.8303
3.4233	2.0	30	2.9354	0.5224	18.8283
3.2873	3.0	45	2.9614	0.5219	19.3255
3.2365	4.0	60	2.8914	0.5229	18.0180
3.1492	5.0	75	2.8724	0.5227	17.6798
3.1389	6.0	90	2.8571	0.5244	17.4115
3.0779	7.0	105	2.9126	0.5161	18.4049
3.0185	8.0	120	2.8483	0.5246	17.2577
3.0545	9.0	135	2.7652	0.5365	15.8829
3.0221	10.0	150	2.8007	0.5275	16.4558
3.021	11.0	165	2.7138	0.5408	15.0869
2.9357	12.0	180	2.7198	0.5405	15.1771
2.9179	13.0	195	2.7700	0.5302	15.9584
2.9423	14.0	210	2.7747	0.5323	16.0344
2.9322	15.0	225	2.6850	0.5456	14.6582
2.9055	16.0	240	2.6724	0.5398	14.4746
2.8717	17.0	255	2.7270	0.5416	15.2866
2.8626	18.0	270	2.7876	0.5343	16.2428
2.8691	19.0	285	2.7013	0.5381	14.8989
2.8442	20.0	300	2.6774	0.5441	14.5477
2.8474	21.0	315	2.8358	0.5250	17.0435

Framework versions

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