meta-llama-Meta-Llama-3.1-8B-Instruct_SFT_E1_D10001

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct model using unsloth.

Model description

The Model is trained on all successful episodes of the clembench-benchmark versions 0.9 and 1.0. The Dataset contains approximately 3700 Successfully player episodes of all non-multi-modal games

Training and evaluation data

Dataset: D10001

Training procedure

One Episode QLoRa Finetuning with 4bit quantization

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 7331
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.03
• lr_scheduler_warmup_steps: 5
• num_epochs: 1

Training results

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1