fzwd6666/NLI_new

This model is a fine-tuned version of bert-base-uncased on an NLI dataset. It achieves the following results on the evaluation set:

{'precision': 0.9690210656753407} {'recall': 0.9722337339411521} {'f1': 0.9706247414149772} {'accuracy': 0.9535340314136126}

Training hyperparameters:

learning_rate: 2e-5

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': 0.623370110330993} {'recall': 0.8415707515233581} {'f1': 0.7162201094785364} {'accuracy': 0.5828038966539602} Incorrect UD Unigram: {'precision': 0.6211431461810825} {'recall': 0.8314150304671631} {'f1': 0.7110596409959468} {'accuracy': 0.5772977551884795} Incorrect UD Bigram: {'precision': 0.6203980099502487} {'recall': 0.8442789438050101} {'f1': 0.7152279896759391} {'accuracy': 0.579415501905972} Incorrect UD All: {'precision': 0.605543710021322} {'recall': 0.1922816519972918} {'f1': 0.2918807810894142} {'accuracy': 0.4163490046590428} Incorrect Sentence: {'precision': 0.6411042944785276} {'recall': 0.4245091401489506} {'f1': 0.5107942973523422} {'accuracy': 0.4913172384582804}