Modelpublic

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

sourceHugging Faceupdated 2y agoView on Hugging Face

0likes6downloads

Model Card

peft-lora-CodeLlama-13b-flutter-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: 0.3620

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
0.7824	0.05	100	0.4186
0.3055	0.1	200	0.4164
0.4455	0.15	300	0.4134
0.3148	0.2	400	0.3762
0.2942	0.25	500	0.3780
0.8817	0.3	600	0.3760
0.4958	0.35	700	0.3738
0.4388	0.4	800	0.3710
0.3605	0.45	900	0.3698
0.2862	0.5	1000	0.3673
3.4798	0.55	1100	0.3687
3.3077	0.6	1200	0.3703
0.4847	0.65	1300	0.3666
0.3593	0.7	1400	0.3662
0.5983	0.75	1500	0.3654
0.6138	0.8	1600	0.3638
0.403	0.85	1700	0.3635
0.4199	0.9	1800	0.3632
0.3526	0.95	1900	0.3621
0.492	1.0	2000	0.3620

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