| license: other | |
| tags: | |
| - generated_from_trainer | |
| datasets: | |
| - ChaiML/dalio_combined_v1 | |
| model-index: | |
| - name: 6.7b-ri-reproduce-combined-4-gpu-0-val | |
| 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. --> | |
| # 6.7b-ri-reproduce-combined-4-gpu-0-val | |
| This model is a fine-tuned version of [facebook/opt-6.7b](https://huggingface.co/facebook/opt-6.7b) on the ChaiML/dalio_combined_v1 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: 9e-07 | |
| - train_batch_size: 1 | |
| - eval_batch_size: 8 | |
| - seed: 100 | |
| - distributed_type: multi-GPU | |
| - num_devices: 4 | |
| - total_train_batch_size: 4 | |
| - total_eval_batch_size: 32 | |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 | |
| - lr_scheduler_type: constant | |
| - num_epochs: 15.0 | |
| ### Training results | |
| ### Framework versions | |
| - Transformers 4.25.0.dev0 | |
| - Pytorch 1.12.1+cu113 | |
| - Datasets 2.3.2 | |
| - Tokenizers 0.12.1 | |