Modelpublic

vasugupta0607/llama2_instruct_generation

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes6downloads

Model Card

llama2instructgeneration

This model is a fine-tuned version of NousResearch/Llama-2-7b-hf on the generator dataset. It achieves the following results on the evaluation set:

—Loss: 1.6734

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: 0.0002
—trainbatchsize: 4
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: constant
—lrschedulerwarmup_steps: 0.03
—training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss
1.9364	0.0	20	1.8092
1.9198	0.01	40	1.7826
1.8451	0.01	60	1.7675
1.8487	0.01	80	1.7573
1.8667	0.01	100	1.7435
1.7463	0.02	120	1.7132
1.7789	0.02	140	1.7038
1.8167	0.02	160	1.7008
1.8654	0.02	180	1.6944
1.9158	0.03	200	1.6939
1.6581	0.03	220	1.6909
1.793	0.03	240	1.6896
1.7878	0.04	260	1.6872
1.7542	0.04	280	1.6862
1.7723	0.04	300	1.6863
1.7606	0.04	320	1.6832
1.8054	0.05	340	1.6802
1.7307	0.05	360	1.6803
1.8278	0.05	380	1.6790
1.7912	0.05	400	1.6768
1.7826	0.06	420	1.6749
1.8975	0.06	440	1.6756
1.8395	0.06	460	1.6763
1.8319	0.07	480	1.6749
1.7879	0.07	500	1.6734

Framework versions

—PEFT 0.7.1
—Transformers 4.37.2
—Pytorch 2.1.0+cu121
—Datasets 2.16.1
—Tokenizers 0.15.1