| tags: | |
| - summarization | |
| - en | |
| - ecnoder-decoder | |
| - xlmroberta | |
| - Abstractive Summarization | |
| - generated_from_trainer | |
| datasets: | |
| - cnn_dailymail | |
| model-index: | |
| - name: xlmroberta-finetune-en-cnn | |
| results: [] | |
| <!-- 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. --> | |
| # xlmroberta-finetune-en-cnn | |
| This model is a fine-tuned version of [](https://huggingface.co/) on the cnn_dailymail dataset. | |
| ## 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: 0.0005 | |
| - train_batch_size: 4 | |
| - eval_batch_size: 4 | |
| - seed: 42 | |
| - gradient_accumulation_steps: 8 | |
| - total_train_batch_size: 32 | |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 | |
| - lr_scheduler_type: linear | |
| - lr_scheduler_warmup_steps: 250 | |
| - num_epochs: 5 | |
| - label_smoothing_factor: 0.1 | |
| ### Framework versions | |
| - Transformers 4.19.4 | |
| - Pytorch 1.11.0+cu113 | |
| - Datasets 2.3.0 | |
| - Tokenizers 0.12.1 | |