mt5-base-wikinewssum-french

This model is a fine-tuned version of google/mt5-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0917
• Rouge1: 12.0984
• Rouge2: 5.7289
• Rougel: 9.9245
• Rougelsum: 11.0697

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: 5.6e-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: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
No log	1.0	549	2.3203	11.5172	4.9352	9.3617	10.4605
No log	2.0	1098	2.2057	11.8469	5.2369	9.6452	10.8337
No log	3.0	1647	2.1525	11.9096	5.4027	9.7648	10.9315
3.1825	4.0	2196	2.1307	12.0782	5.5848	9.9614	11.1081
3.1825	5.0	2745	2.1172	11.9821	5.6042	9.8216	11.0077
3.1825	6.0	3294	2.1012	12.0845	5.6834	9.9119	11.0741
3.1825	7.0	3843	2.0964	12.1296	5.7271	9.9495	11.1227
2.3376	8.0	4392	2.0917	12.0984	5.7289	9.9245	11.0697

Framework versions

• Transformers 4.13.0
• Pytorch 1.10.1
• Datasets 1.16.1
• Tokenizers 0.10.3