long-t5-tglobal-large-pubmed-3k-booksum-16384-WIP13
Evaluating some metric results before merging with the "main" wip version
This model is a fine-tuned version of pszemraj/long-t5-tglobal-large-pubmed-3k-booksum-16384-WIP12 on the kmfoda/booksum
.
The "base" checkpoint that I update when a training session is productive is here
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: 0.0006
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 64
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.01
- num_epochs: 1.1
Framework versions
- Transformers 4.21.2
- Pytorch 1.10.0+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1
- Downloads last month
- 4
Model tree for pszemraj/long-t5-tglobal-large-pubmed-3k-booksum-16384-WIP13
Dataset used to train pszemraj/long-t5-tglobal-large-pubmed-3k-booksum-16384-WIP13
Evaluation results
- ROUGE-1 on samsumtest set verified24.410
- ROUGE-2 on samsumtest set verified5.003
- ROUGE-L on samsumtest set verified17.254
- ROUGE-LSUM on samsumtest set verified20.918
- loss on samsumtest set verified3.195
- gen_len on samsumtest set verified58.995
- ROUGE-1 on billsumtest set verified37.365
- ROUGE-2 on billsumtest set verified12.332
- ROUGE-L on billsumtest set verified22.075
- ROUGE-LSUM on billsumtest set verified31.168