README.md

metadata

license: mit
tags:
  - generated_from_trainer
model-index:
  - name: roberta-base-mnli_CollSgE
    results: []

roberta-base-mnli_CollSgE

This model is a fine-tuned version of WillHeld/roberta-base-mnli on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7610
• Acc: 0.8445

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Acc
0.4123	0.17	2000	0.4693	0.8332
0.4028	0.33	4000	0.4624	0.8338
0.3888	0.5	6000	0.4500	0.8375
0.3841	0.67	8000	0.4281	0.8416
0.3783	0.83	10000	0.4434	0.8365
0.3759	1.0	12000	0.4400	0.8418
0.2721	1.17	14000	0.5022	0.8427
0.2736	1.33	16000	0.5252	0.8431
0.2821	1.5	18000	0.4887	0.8409
0.2802	1.67	20000	0.4758	0.8458
0.2794	1.83	22000	0.4611	0.8458
0.2797	2.0	24000	0.4936	0.8456
0.1915	2.17	26000	0.5545	0.8462
0.1946	2.33	28000	0.5731	0.8443
0.2007	2.5	30000	0.5507	0.8428
0.2008	2.67	32000	0.5499	0.8454
0.1971	2.84	34000	0.5274	0.8483
0.2054	3.0	36000	0.5454	0.8476
0.1436	3.17	38000	0.6787	0.8442
0.1426	3.34	40000	0.6933	0.8421
0.1463	3.5	42000	0.6547	0.8455
0.1447	3.67	44000	0.6469	0.8438
0.1445	3.84	46000	0.6626	0.8472
0.1457	4.0	48000	0.6494	0.8504
0.1133	4.17	50000	0.7664	0.8459
0.1138	4.34	52000	0.7857	0.8452
0.1154	4.5	54000	0.7623	0.8486
0.1102	4.67	56000	0.7740	0.8460
0.1143	4.84	58000	0.7610	0.8445

Framework versions

• Transformers 4.24.0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1
• Tokenizers 0.12.1