llama_8b_lima_42

This model is a fine-tuned version of meta-llama/Llama-3.1-8B on the open_webui_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.9014

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: 5e-06
• train_batch_size: 3
• eval_batch_size: 2
• seed: 66
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 6
• total_train_batch_size: 36
• total_eval_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: polynomial
• lr_scheduler_warmup_steps: 50
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.0356	0.0686	80	0.9711
1.0348	0.1372	160	0.9355
0.7212	0.2058	240	0.9225
0.8077	0.2744	320	0.9142
0.8442	0.3430	400	0.9068
0.831	0.4116	480	0.9032
0.9696	0.4802	560	0.8975
0.9949	0.5488	640	0.8972
0.8154	0.6174	720	0.8948
0.9682	0.6860	800	0.8931
0.9491	0.7546	880	0.9030
1.068	0.8232	960	0.9022
1.06	0.8918	1040	0.9013
0.8302	0.9604	1120	0.9002

Framework versions

• Transformers 4.46.1
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3