metadata
license: mit
base_model: TheBloke/zephyr-7B-alpha-GPTQ
tags:
- generated_from_trainer
model-index:
- name: latest
results: []
latest
This model is a fine-tuned version of TheBloke/zephyr-7B-alpha-GPTQ on the None 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: 0.0002
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- training_steps: 50
- mixed_precision_training: Native AMP
Training results
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu118
- Datasets 2.15.0
- Tokenizers 0.15.0