griffin-1024-llama3t-8layer-simple_wikipedia_LM-vN
pretraining experiment on the pszemraj/simple_wikipedia_LM dataset. It achieves the following results on the evaluation set:
- Loss: 4.3584
- Accuracy: 0.3789
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: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 80085
- gradient_accumulation_steps: 32
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-07
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 2.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 13.6044 | 0.2495 | 100 | 12.5441 | 0.0079 |
| 8.9524 | 0.4989 | 200 | 8.4254 | 0.0473 |
| 7.1721 | 0.7484 | 300 | 6.6199 | 0.0389 |
| 6.2087 | 0.9978 | 400 | 5.7198 | 0.2251 |
| 5.4917 | 1.2473 | 500 | 4.9480 | 0.3268 |
| 4.9408 | 1.4967 | 600 | 4.6730 | 0.3567 |
| 4.8347 | 1.7462 | 700 | 4.4984 | 0.3707 |
| 4.7023 | 1.9956 | 800 | 4.3584 | 0.3789 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1