Modelpublic

KaQyn/peft-lora-CodeLlama-13b-unity-copilot

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes5downloads

Model Card

peft-lora-CodeLlama-13b-unity-copilot

This model is a fine-tuned version of codellama/CodeLlama-13b-Instruct-hf on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.1825

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: 0.0003
—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_ratio: 0.1
—training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss
0.1196	0.05	100	0.3165
0.2451	0.1	200	0.2333
0.2358	0.15	300	0.2338
0.1617	0.2	400	0.2239
0.1839	0.25	500	0.2154
0.1845	0.3	600	0.2169
0.2536	0.35	700	0.2185
0.1649	0.4	800	0.2153
0.2767	0.45	900	0.2056
0.1939	0.5	1000	0.2120
0.1093	0.55	1100	0.1975
0.1852	0.6	1200	0.1912
0.1077	0.65	1300	0.1930
0.1129	0.7	1400	0.1959
0.1219	0.75	1500	0.1958
0.3024	0.8	1600	0.1882
0.2656	0.85	1700	0.1872
0.1666	0.9	1800	0.1849
0.1553	0.95	1900	0.1827
0.1477	1.0	2000	0.1825

Framework versions

—PEFT 0.10.1.dev0
—Transformers 4.40.0.dev0
—Pytorch 2.2.1+cu121
—Datasets 2.18.0
—Tokenizers 0.15.2