german-jeopardy-longt5-base-128

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

german-jeopardy-longt5-base-128

This model is a fine-tuned version of google/long-t5-tglobal-base on the lmqg/qg_dequad dataset. It achieves the following results on the evaluation set:

• Loss: 1.8010
• Brevity Penalty: 0.8577
• System Length: 18026
• Reference Length: 20793
• ROUGE-1: 35.34
• ROUGE-2: 16.82
• ROUGE-L: 34.13
• ROUGE-Lsum: 34.14
• Exact Match: 1.41
• BLEU: 10.73
• F1: 34.55

Model description

See google/long-t5-tglobal-base for more information about the model architecture.
The model was trained on a single NVIDIA RTX 3090 GPU with 24GB of VRAM.

Intended uses & limitations

This model can be used for question generation on German text.

Training and evaluation data

See lmqg/qg_dequad.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 8
• eval_batch_size: 4
• seed: 7
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adafactor
• lr_scheduler_type: constant
• num_epochs: 20

Training results

Framework versions

• Transformers 4.32.1
• Pytorch 2.1.0
• Datasets 2.12.0
• Tokenizers 0.13.3