light-prompt-compression-multilingual

This model is a fine-tuned version of google/mt5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8427
• Rouge1: 0.7961
• Rouge2: 0.6052
• Rougel: 0.7532
• Rougelsum: 0.7531
• Comp Ratio Mean: 0.5926
• Comp Ratio P90: 0.75
• Pct Violations: 0.0

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAFACTOR and the args are: No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 7

Training results

Training Loss	Epoch	Step	Comp Ratio Mean	Comp Ratio P90	Validation Loss	Pct Violations	Rouge1	Rouge2	Rougel	Rougelsum
8.3491	1.0	3996	0.6193	0.7647	1.4078	0.0002	0.7243	0.5233	0.6801	0.6803
1.728	2.0	7992	0.6033	0.7619	1.0550	0.0	0.7699	0.5722	0.7247	0.7246
1.3959	3.0	11988	0.5972	0.75	0.9433	0.0	0.7841	0.5889	0.7403	0.7401
1.2622	4.0	15984	0.5925	0.75	0.8947	0.0	0.7910	0.5973	0.7473	0.7471
1.1867	5.0	19980	0.5911	0.75	0.8567	0.0	0.7945	0.6019	0.7511	0.7511
1.1602	6.0	23976	0.5919	0.75	0.8516	0.0	0.7957	0.6043	0.7523	0.7523
1.1468	7.0	27972	0.5926	0.75	0.8427	0.0	0.7961	0.6052	0.7532	0.7531

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1