math_question_grade_detection

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

• Loss: 0.5515
• Accuracy: 0.8248
• Precision: 0.8311
• Recall: 0.8248
• F1: 0.8240

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 850

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	0.2732	50	1.4704	0.4704	0.4864	0.4704	0.4383
No log	0.5464	100	1.1159	0.5742	0.5936	0.5742	0.5676
No log	0.8197	150	0.9276	0.6441	0.6564	0.6441	0.6391
No log	1.0929	200	0.7966	0.7064	0.7146	0.7064	0.7057
No log	1.3661	250	0.7308	0.7317	0.7408	0.7317	0.7291
No log	1.6393	300	0.6640	0.7571	0.7624	0.7571	0.7560
No log	1.9126	350	0.5874	0.7940	0.7975	0.7940	0.7931
No log	2.1858	400	0.6288	0.7863	0.7958	0.7863	0.7840
No log	2.4590	450	0.5621	0.8055	0.8128	0.8055	0.8048
0.8255	2.7322	500	0.5799	0.8094	0.8181	0.8094	0.8087
0.8255	3.0055	550	0.5560	0.7994	0.8041	0.7994	0.7978
0.8255	3.2787	600	0.5402	0.8301	0.8341	0.8301	0.8305
0.8255	3.5519	650	0.5534	0.8201	0.8287	0.8201	0.8197
0.8255	3.8251	700	0.5439	0.8248	0.8333	0.8248	0.8249
0.8255	4.0984	750	0.5402	0.8248	0.8304	0.8248	0.8244
0.8255	4.3716	800	0.5363	0.8271	0.8309	0.8271	0.8262
0.8255	4.6448	850	0.5515	0.8248	0.8311	0.8248	0.8240

Framework versions

• Transformers 4.47.0
• Pytorch 2.2.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0