stablelm-2-1_6b-indo-lora

This model is a fine-tuned version of stabilityai/stablelm-2-1_6b on the generator 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: 0.0005
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Framework versions

• PEFT 0.7.1
• Transformers 4.37.0
• Pytorch 2.0.0
• Datasets 2.16.1
• Tokenizers 0.15.0