test-dialogue-summarization

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

• Loss: 1.8387
• Rouge1: 48.1775
• Rouge2: 24.5925
• Rougel: 40.3237
• Rougelsum: 43.9647
• Gen Len: 18.4707

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.8408	1.0	1841	1.5902	47.4895	24.7763	40.0228	44.4895	18.5159
1.5348	2.0	3683	1.5498	48.0242	24.8392	40.559	44.2542	17.6015
1.3076	3.0	5524	1.5561	48.5695	25.9259	41.4698	44.6406	17.4658
1.1286	4.0	7366	1.5796	48.5079	25.1521	40.8084	44.6149	18.4364
0.9956	5.0	9207	1.6134	49.1351	25.6367	41.3139	45.0814	18.3313
0.8668	6.0	11049	1.6679	49.002	25.4589	41.1276	44.787	18.4853
0.7696	7.0	12890	1.7327	48.1978	25.0238	40.6671	44.3866	18.3374
0.69	8.0	14732	1.7603	48.7522	25.0831	40.8193	44.4452	18.4597
0.6175	9.0	16573	1.8092	48.2747	24.8563	40.3027	44.1975	18.3729
0.5701	10.0	18410	1.8387	48.1775	24.5925	40.3237	43.9647	18.4707

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2