mt5-small-finetuned-Big-Patent-h

<!-- 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. -->

mt5-small-finetuned-Big-Patent-h

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

• Loss: 2.2622
• Rouge1: 33.9091
• Rouge2: 14.1731
• Rougel: 30.105
• Rougelsum: 30.3666

Model description

In this project, we fine-tuned mT5small, a multilingual variant of T5 that was pre-trained on a new Common Crawl-based dataset covering 101 languages. The model was fine-tuned on the electric patent corpus using a variety of techniques, including transfer learning, data augmentation, and hyperparameter tuning.

Intended uses & limitations

The fine-tuned model showed significant improvements in performance on the electric patent-specific tasks compared to the original pre-trained model.

Note: This project is suitable for researchers who are working on electric patent, as it's fine-tuned on electric patents and it can be used for related NLP problems for electric patent and electric patent research.

Training and evaluation data

A subset of electric patents were used to fine-tune the model.

The fine-tuned model was evaluated using the ROUGE metric on a variety of natural language processing tasks specific to the patent domain, including, named entity recognition, and summarization.

Training procedure

The model was fine-tuned on the electric patent corpus using a variety of techniques, including transfer learning, data augmentation, and hyperparameter tuning.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5.6e-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: 8

Training results

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2