This model was created for classification users requests for tech support. Users ask the same questions for tech support and operators waste a lot of time answering them. There were instructions, but users had trouble finding them, so we made the decision to make a chat interface where user can text his problem and a pretrained model will classify his question and respond with correct instruction.

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: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	F1
No log	1.0	7	1.7976	0.3806
No log	2.0	14	1.6849	0.3806
No log	3.0	21	1.5963	0.3806
No log	4.0	28	1.4947	0.3806
No log	5.0	35	1.4645	0.3806
No log	6.0	42	1.4063	0.3806
No log	7.0	49	1.4314	0.4935
No log	8.0	56	1.2979	0.5274
No log	9.0	63	1.3582	0.4626
No log	10.0	70	1.5711	0.5164
No log	11.0	77	1.2483	0.5881
No log	12.0	84	1.1974	0.5860
No log	13.0	91	1.2582	0.5426
No log	14.0	98	1.7688	0.4504
No log	15.0	105	1.3278	0.5557
No log	16.0	112	1.6230	0.5119
No log	17.0	119	1.4229	0.5536
No log	18.0	126	1.4000	0.5536
No log	19.0	133	1.4614	0.5408
No log	20.0	140	1.4676	0.5536
No log	21.0	147	1.7174	0.555
No log	22.0	154	1.5338	0.5536
No log	23.0	161	1.6979	0.6179
No log	24.0	168	1.7075	0.5408
No log	25.0	175	1.6655	0.5408
No log	26.0	182	1.6043	0.6179
No log	27.0	189	1.6945	0.6051
No log	28.0	196	1.7289	0.5408
1.1079	29.0	203	1.7329	0.5408
1.1079	30.0	210	1.7348	0.5408

Framework versions

Transformers 4.27.3
Pytorch 2.0.0+cu117
Datasets 2.11.0
Tokenizers 0.13.2

DonMakar/bert-base-Daichi_support

bert-base-Daichi_support

Model description