llama_8b_lima_31

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: 0.9656

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: 6e-06
• train_batch_size: 3
• eval_batch_size: 2
• seed: 66
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 3
• total_train_batch_size: 18
• 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: 40
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.1455	0.0537	40	1.0655
1.0396	0.1073	80	1.0346
0.9381	0.1610	120	1.0174
0.9677	0.2147	160	1.0032
1.0119	0.2683	200	1.0007
1.053	0.3220	240	0.9904
1.1013	0.3757	280	0.9832
0.8192	0.4293	320	0.9825
1.0043	0.4830	360	0.9794
1.0733	0.5367	400	0.9717
0.9807	0.5903	440	0.9713
0.8318	0.6440	480	0.9720
0.9004	0.6977	520	0.9682
1.0613	0.7513	560	0.9656
1.165	0.8050	600	0.9679
0.846	0.8587	640	0.9680
0.9019	0.9123	680	0.9661
0.9271	0.9660	720	0.9649

Framework versions

• Transformers 4.46.1
• Pytorch 2.4.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3