metadata
tags:
- generated_from_trainer
model_index:
- name: hybrid_hbh_t5-small_ami_sum
results:
- task:
name: Summarization
type: summarization
hybrid_hbh_t5-small_ami_sum
This model is a fine-tuned version of t5-small on an AMI dataset for dialogue summarization task.
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
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 100.0
- label_smoothing_factor: 0.1
Results on Test Set
- predict_gen_len = 329.2
- predict_rouge1 = 48.7673
- predict_rouge2 = 18.1832
- predict_rougeL = 26.1713
- predict_rougeLsum = 46.8434
- predict_samples = 20
- predict_samples_per_second = 1.098
- predict_steps_per_second = 0.274
Framework versions
- Transformers 4.8.0
- Pytorch 1.6.0
- Datasets 1.10.2
- Tokenizers 0.10.3