mt5-base-wikinewssum-portuguese
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Base Model | google/mt5-base |
| Framework | PyTorch 1.10.1 |
| Task | Text Summarization |
What is mt5-base-wikinewssum-portuguese?
mt5-base-wikinewssum-portuguese is a specialized text summarization model built on Google's mT5 architecture, specifically fine-tuned for Portuguese language content. The model demonstrates promising performance with a ROUGE-1 score of 9.49 and ROUGE-2 score of 4.22, making it particularly suitable for Portuguese text summarization tasks.
Implementation Details
The model was trained using a carefully optimized process with the following specifications: learning rate of 5.6e-05, batch size of 8 (with gradient accumulation), and Adam optimizer. The training spanned 8 epochs with a linear learning rate scheduler, showing consistent improvement in ROUGE scores throughout the training process.
- Transformers version: 4.13.0
- Gradient accumulation steps: 2
- Total training epochs: 8
- Final validation loss: 2.0428
Core Capabilities
- Portuguese text summarization
- Multi-lingual text processing foundation (mT5)
- Optimized for news content summarization
- Consistent performance across different ROUGE metrics
Frequently Asked Questions
Q: What makes this model unique?
This model's uniqueness lies in its specialized fine-tuning for Portuguese news summarization, built on the robust mT5 architecture, making it particularly effective for Portuguese language content processing.
Q: What are the recommended use cases?
The model is best suited for Portuguese news article summarization, content condensation, and general text summarization tasks where maintaining key information while reducing text length is crucial.
