flan-t5-base-samsum

This model is a fine-tuned version of google/flan-t5-base on the samsum dataset a collection of about 16k messenger-like conversations with summaries. Conversations were created and written down by linguists fluent in English. It achieves the following results on the evaluation set:

• Loss: 1.3715
• Rouge1: 47.339
• Rouge2: 23.8991
• Rougel: 40.0668
• Rougelsum: 43.6669
• Gen Len: 17.2063

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: 8
• eval_batch_size: 8
• 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	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.4525	1.0	1842	1.3841	46.331	22.8895	39.0677	42.845	17.2149
1.3436	2.0	3684	1.3732	47.0523	23.5437	39.8169	43.449	17.1954
1.2821	3.0	5526	1.3717	47.2418	23.6518	39.774	43.5104	17.2295
1.2307	4.0	7368	1.3715	47.339	23.8991	40.0668	43.6669	17.2063
1.1985	5.0	9210	1.3770	47.4925	24.0124	40.128	43.8232	17.2833

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1