financial-twhin-bert-large-3labels-test1

This model is a fine-tuned version of Twitter/twhin-bert-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3334
• Accuracy: 0.8826
• F1: 0.8823

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: 9.656814753771254e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 1203
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.9822	0.1550	100	0.7065	0.6772	0.5469
0.7307	0.3101	200	0.5716	0.7471	0.7179
0.6482	0.4651	300	0.5388	0.7716	0.7493
0.6008	0.6202	400	0.4300	0.8494	0.8446
0.5237	0.7752	500	0.4190	0.8343	0.8401
0.5106	0.9302	600	0.4114	0.8444	0.8404
0.4832	1.0853	700	0.3865	0.8545	0.8596
0.4031	1.2403	800	0.3741	0.8602	0.8653
0.3729	1.3953	900	0.3334	0.8826	0.8823
0.3661	1.5504	1000	0.3494	0.8725	0.8750
0.332	1.7054	1100	0.3390	0.8725	0.8753
0.3637	1.8605	1200	0.3386	0.8689	0.8724

Framework versions

• Transformers 4.40.0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1