train_hellaswag_42_1760637624

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the hellaswag dataset. It achieves the following results on the evaluation set:

• Loss: 0.7598
• Num Input Tokens Seen: 193996096

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.4586	2.0	15962	0.4638	19409344
0.4621	4.0	31924	0.4639	38802816
0.4707	6.0	47886	0.4626	58213312
0.4615	8.0	63848	0.4641	77597952
0.4544	10.0	79810	0.4662	96994976
0.4884	12.0	95772	0.4926	116401696
0.3192	14.0	111734	0.5258	135803712
0.3558	16.0	127696	0.6282	155205696
0.461	18.0	143658	0.7163	174600352
0.1986	20.0	159620	0.7598	193996096

Framework versions

• PEFT 0.17.1
• Transformers 4.51.3
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4