Modelpublic

ZiHDeng/peft-lora-starcoder1B-Instruction-ny8-ALL

sourceHugging Facebigcode-openrail-mupdated 3y agoView on Hugging Face

0likes19downloads

Model Card

peft-lora-starcoder1B-Instruction-ny8-ALL

This model is a fine-tuned version of bigcode/starcoderbase-1b on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.0870

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: 0.0005
—trainbatchsize: 4
—evalbatchsize: 4
—seed: 42
—gradientaccumulationsteps: 4
—totaltrainbatch_size: 16
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 30
—training_steps: 2000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.1891	0.05	100	0.1452
0.1244	0.1	200	0.1096
0.1077	0.15	300	0.1006
0.0996	0.2	400	0.0958
0.0953	0.25	500	0.0927
0.0916	0.3	600	0.0882
0.0875	0.35	700	0.0867
0.0845	0.4	800	0.0873
0.0818	0.45	900	0.0863
0.0788	0.5	1000	0.0848
0.0781	0.55	1100	0.0844
0.0749	0.6	1200	0.0847
0.0726	0.65	1300	0.0849
0.0688	0.7	1400	0.0867
0.0701	0.75	1500	0.0861
0.0662	0.8	1600	0.0863
0.0658	0.85	1700	0.0867
0.0647	0.9	1800	0.0869
0.0644	0.95	1900	0.0870
0.0657	1.0	2000	0.0870

Framework versions

—PEFT 0.7.1
—Transformers 4.36.2
—Pytorch 2.1.2+cu121
—Datasets 2.16.1
—Tokenizers 0.15.0