newsdiscourse-model-large

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

• Loss: 1.5899
• F1: 0.1975

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: 3e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5.0

Training results

Training Loss	Epoch	Step	Validation Loss	F1
No log	0.14	100	1.9895	0.0487
No log	0.28	200	2.0130	0.0512
No log	0.43	300	1.9527	0.0512
No log	0.57	400	1.9605	0.0487
2.0539	0.71	500	1.9854	0.0618
2.0539	0.85	600	1.7978	0.1242
2.0539	1.0	700	1.7291	0.1373
2.0539	1.14	800	1.9082	0.0487
2.0539	1.28	900	1.9300	0.0487
1.9096	1.42	1000	1.7186	0.1414
1.9096	1.57	1100	1.7304	0.1399
1.9096	1.71	1200	1.7281	0.1363
1.9096	1.85	1300	1.8452	0.0576
1.9096	1.99	1400	1.7180	0.1519
1.7842	2.14	1500	1.7450	0.1525
1.7842	2.28	1600	1.7752	0.1344
1.7842	2.42	1700	1.7548	0.1506
1.7842	2.56	1800	1.7185	0.1536
1.7842	2.71	1900	1.6870	0.1536
1.7227	2.85	2000	1.7336	0.1536
1.7227	2.99	2100	1.7217	0.1490
1.7227	3.13	2200	1.7213	0.1482
1.7227	3.28	2300	1.7482	0.1435
1.7227	3.42	2400	1.7559	0.1456
1.7441	3.56	2500	1.7324	0.1406
1.7441	3.7	2600	1.6977	0.1484
1.7441	3.85	2700	1.6276	0.1839
1.7441	3.99	2800	1.6109	0.1876
1.7441	4.13	2900	1.6359	0.2181
1.6515	4.27	3000	1.6463	0.1792
1.6515	4.42	3100	1.6397	0.1828
1.6515	4.56	3200	1.6189	0.1837
1.6515	4.7	3300	1.6096	0.1875
1.6515	4.84	3400	1.5904	0.1925
1.6003	4.99	3500	1.5899	0.1975

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3