Edit model card

SGH logo.png This model is a fine-tuned version of allenai/led-base-16384 on the multi_news dataset. It achieves the following results on the evaluation set:

  • Loss: 2.3650
  • Rouge1 Precision: 0.4673
  • Rouge1 Recall: 0.4135
  • Rouge1 Fmeasure: 0.4263
  • Rouge2 Precision: 0.1579
  • Rouge2 Recall: 0.1426
  • Rouge2 Fmeasure: 0.1458
  • Rougel Precision: 0.2245
  • Rougel Recall: 0.2008
  • Rougel Fmeasure: 0.2061
  • Rougelsum Precision: 0.2245
  • Rougelsum Recall: 0.2008
  • Rougelsum Fmeasure: 0.2061

Model description

This model was created to generate summaries of news articles.

Intended uses & limitations

The model takes up to maximum article length of 3072 tokens and generates a summary of maximum length of 512 tokens, and minimum length of 100 tokens.

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: 4
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Rouge1 Precision Rouge1 Recall Rouge1 Fmeasure Rouge2 Precision Rouge2 Recall Rouge2 Fmeasure Rougel Precision Rougel Recall Rougel Fmeasure Rougelsum Precision Rougelsum Recall Rougelsum Fmeasure
2.8095 0.16 10 2.5393 0.287 0.5358 0.3674 0.1023 0.1917 0.1311 0.1374 0.2615 0.1771 0.1374 0.2615 0.1771
2.6056 0.32 20 2.4752 0.5005 0.3264 0.3811 0.1663 0.1054 0.1249 0.2582 0.1667 0.1957 0.2582 0.1667 0.1957
2.5943 0.48 30 2.4422 0.4615 0.3833 0.4047 0.1473 0.1273 0.1321 0.2242 0.1885 0.1981 0.2242 0.1885 0.1981
2.4842 0.64 40 2.4186 0.4675 0.3829 0.4081 0.1581 0.1294 0.1384 0.2286 0.187 0.1995 0.2286 0.187 0.1995
2.4454 0.8 50 2.3990 0.467 0.408 0.4222 0.1633 0.1429 0.1477 0.2294 0.2008 0.2076 0.2294 0.2008 0.2076
2.3622 0.96 60 2.3857 0.4567 0.3898 0.41 0.1433 0.1233 0.1295 0.2205 0.1876 0.1976 0.2205 0.1876 0.1976
2.4034 1.13 70 2.3835 0.4515 0.4304 0.4294 0.1526 0.1479 0.1459 0.2183 0.209 0.2078 0.2183 0.209 0.2078
2.2612 1.29 80 2.3804 0.455 0.4193 0.4236 0.1518 0.1429 0.1427 0.2177 0.2025 0.2037 0.2177 0.2025 0.2037
2.2563 1.45 90 2.3768 0.4821 0.391 0.4196 0.1652 0.1357 0.144 0.2385 0.1929 0.2069 0.2385 0.1929 0.2069
2.243 1.61 100 2.3768 0.4546 0.4093 0.4161 0.1552 0.1402 0.1422 0.2248 0.2016 0.2052 0.2248 0.2016 0.2052
2.2505 1.77 110 2.3670 0.4625 0.4189 0.4262 0.1606 0.1485 0.1493 0.2301 0.2098 0.2119 0.2301 0.2098 0.2119
2.2453 1.93 120 2.3650 0.4673 0.4135 0.4263 0.1579 0.1426 0.1458 0.2245 0.2008 0.2061 0.2245 0.2008 0.2061

Framework versions

  • Transformers 4.21.3
  • Pytorch 1.12.1+cu113
  • Datasets 2.6.2.dev0
  • Tokenizers 0.12.1
Downloads last month
9
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.

Dataset used to train debbiesoon/summarise_v9