llama-3-8b-finetuned-turkish-instructions

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on 3641 Turkish instructions in the aya_dataset.

Training procedure

This model is fine-tuned using QloRA and SFT.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 3
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 6
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 5

Framework versions

• PEFT 0.10.1.dev0
• Transformers 4.41.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1