Edit model card

text_shortening_model_v40

This model is a fine-tuned version of facebook/bart-large-xsum on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 3.3335
  • Rouge1: 0.4511
  • Rouge2: 0.2377
  • Rougel: 0.4039
  • Rougelsum: 0.4038
  • Bert precision: 0.8635
  • Bert recall: 0.8629
  • Average word count: 8.5826
  • Max word count: 16
  • Min word count: 5
  • Average token count: 16.5616
  • % shortened texts with length > 12: 4.8048

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: 0.0003
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Bert precision Bert recall Average word count Max word count Min word count Average token count % shortened texts with length > 12
3.0922 1.0 73 2.2144 0.4539 0.2272 0.4068 0.4055 0.8657 0.8684 8.7027 15 5 14.3423 4.2042
1.75 2.0 146 2.0055 0.4658 0.2381 0.4085 0.4088 0.8654 0.8656 8.7087 16 5 15.1652 4.8048
1.311 3.0 219 2.0021 0.456 0.2257 0.4124 0.4117 0.8644 0.8646 8.6396 15 5 15.9279 5.1051
1.0163 4.0 292 2.0698 0.467 0.2403 0.4159 0.4162 0.8636 0.8699 9.2973 16 5 17.2162 9.9099
0.8546 5.0 365 2.0707 0.4527 0.2392 0.4129 0.4126 0.8637 0.8647 8.4895 17 4 16.3153 4.8048
0.7222 6.0 438 2.1452 0.4562 0.2349 0.4077 0.4064 0.8693 0.8623 8.021 15 4 14.1051 1.2012
0.5723 7.0 511 2.3520 0.4563 0.2403 0.4142 0.413 0.8666 0.8658 8.5916 16 5 16.5465 6.9069
0.5274 8.0 584 2.2896 0.4502 0.2434 0.4077 0.4078 0.8639 0.8639 8.5586 14 5 14.8048 2.1021
0.3767 9.0 657 2.2928 0.4565 0.2368 0.4125 0.4114 0.8682 0.8623 8.0691 14 4 14.4204 1.8018
0.2987 10.0 730 2.5411 0.4539 0.2383 0.4057 0.4056 0.8652 0.8631 8.5826 15 5 15.6637 4.5045
0.2319 11.0 803 2.8995 0.4513 0.2367 0.4069 0.4068 0.8631 0.8622 8.6607 17 5 16.4535 5.7057
0.2167 12.0 876 2.7950 0.4632 0.2521 0.4163 0.4162 0.8673 0.8679 8.7267 16 4 16.3243 6.3063
0.1952 13.0 949 2.6240 0.4537 0.2396 0.406 0.4059 0.8632 0.8648 8.8258 18 5 16.2613 7.8078
0.1395 14.0 1022 2.8894 0.4588 0.2412 0.4141 0.4144 0.864 0.8658 8.6216 15 5 16.6426 3.6036
0.1298 15.0 1095 2.7580 0.4562 0.2384 0.4085 0.4088 0.8661 0.8659 8.5586 15 5 16.3634 5.4054
0.1044 16.0 1168 2.7724 0.466 0.2527 0.4175 0.4171 0.8677 0.8694 8.7387 15 4 16.4535 5.1051
0.0944 17.0 1241 2.9161 0.4429 0.232 0.3986 0.3986 0.8619 0.8621 8.6306 16 5 16.5255 5.4054
0.077 18.0 1314 3.1718 0.4549 0.2372 0.4054 0.4052 0.863 0.8639 8.6456 15 5 16.7447 4.8048
0.0561 19.0 1387 3.2650 0.4581 0.2413 0.4092 0.4089 0.866 0.865 8.5195 16 5 16.4174 4.8048
0.0542 20.0 1460 3.3335 0.4511 0.2377 0.4039 0.4038 0.8635 0.8629 8.5826 16 5 16.5616 4.8048

Framework versions

  • Transformers 4.33.1
  • Pytorch 2.0.1+cu118
  • Datasets 2.14.5
  • Tokenizers 0.13.3
Downloads last month
4
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for ldos/text_shortening_model_v40

Finetuned
(50)
this model