metadata

license: other
base_model: microsoft/Orca-2-13b
tags:
  - generated_from_trainer
model-index:
  - name: qlora
    results: []

qlora

This model is a fine-tuned version of microsoft/Orca-2-13b on a subset of the Vezora/Mini_Orca_Uncencored_Alpaca dataset, with some particularly spicy prompts added as well, to reduce the risk of rejections.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 6
eval_batch_size: 6
seed: 42
gradient_accumulation_steps: 6
total_train_batch_size: 36
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 15
num_epochs: 2