bert-large-cased-lora-finetuned-ner-EMBO-SourceData

This model is a fine-tuned version of bert-large-cased.

It achieves the following results on the evaluation set:

• Loss: 0.1282
• Precision: 0.7999
• Recall: 0.8278
• F1: 0.8136
• Accuracy: 0.9584

Model description

For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Token%20Classification/Monolingual/EMBO-SourceData%20with%20LoRA/NER%20Project%20Using%20EMBO-SourceData%20with%20LoRA.ipynb

Intended uses & limitations

This model is intended to demonstrate my ability to solve a complex problem using technology.

Training and evaluation data

Dataset Source: https://huggingface.co/datasets/EMBO/BLURB

Token Distribution

Token Distribution After Removing 'O' Tokens

Histogram of Tokenized Input Lengths

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.1552	1.0	3454	0.1499	0.7569	0.7968	0.7763	0.9516
0.1179	2.0	6908	0.1328	0.7910	0.8120	0.8013	0.9564
0.0998	3.0	10362	0.1282	0.7999	0.8278	0.8136	0.9584

Framework versions

• Transformers 4.26.1
• Pytorch 2.0.1
• Datasets 2.13.1
• Tokenizers 0.13.3