fzwd6666/NLTBert_multi_fine_tune_new

This model is a fine-tuned version of fzwd6666/Ged_bert_new with 4 layers on an NLT dataset. It achieves the following results on the evaluation set:

{'precision': 0.9795081967213115} {'recall': 0.989648033126294} {'f1': 0.984552008238929} {'accuracy': 0.9843227424749164}

Training hyperparameters:

learning_rate: 1e-4 train_batch_size: 8 eval_batch_size: 8 optimizer: AdamW with betas=(0.9,0.999) and epsilon=1e-08 weight_decay= 0.01 lr_scheduler_type: linear num_epochs: 3

It achieves the following results on the test set:

Incorrect UD Padded: {'precision': 0.6878048780487804} {'recall': 0.2863913337846987} {'f1': 0.4043977055449331} {'accuracy': 0.4722575180008471}

Incorrect UD Unigram: {'precision': 0.6348314606741573} {'recall': 0.3060257278266757} {'f1': 0.4129739607126542} {'accuracy': 0.4557390936044049}

Incorrect UD Bigram: {'precision': 0.6588419405320813} {'recall': 0.28503723764387273} {'f1': 0.3979206049149338} {'accuracy': 0.4603981363828886}

Incorrect UD All: {'precision': 0.4} {'recall': 0.0013540961408259986} {'f1': 0.002699055330634278} {'accuracy': 0.373994070309191}

Incorrect Sentence: {'precision': 0.5} {'recall': 0.012186865267433988} {'f1': 0.02379378717779247} {'accuracy': 0.37441761965268955}