Modelpublic

Ferdi/CodeLlama-7B-Instruct-AWQ-FaVe-rank16-10epochs

sourceHugging Facellama2updated 2y agoView on Hugging Face

0likes5downloads

Model Card

CodeLlama-7B-Instruct-AWQ-FaVe-rank16-10epochs

This model is a fine-tuned version of TheBloke/CodeLlama-7B-Instruct-AWQ on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.3637

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: 0.0001
—trainbatchsize: 1
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 4
—totaltrainbatch_size: 4
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 10
—num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.2685	10	2.3021
2.3962	0.5369	20	1.9504
2.3962	0.8054	30	1.4355
1.474	1.0738	40	1.0432
1.474	1.3423	50	0.8599
0.8425	1.6107	60	0.7518
0.8425	1.8792	70	0.7018
0.6901	2.1477	80	0.6548
0.6901	2.4161	90	0.6103
0.5934	2.6846	100	0.5735
0.5934	2.9530	110	0.5495
0.5161	3.2215	120	0.5313
0.5161	3.4899	130	0.5076
0.5026	3.7584	140	0.4888
0.5026	4.0268	150	0.4601
0.4304	4.2953	160	0.4383
0.4304	4.5638	170	0.4298
0.3778	4.8322	180	0.4311
0.3778	5.1007	190	0.4134
0.3189	5.3691	200	0.4027
0.3189	5.6376	210	0.3946
0.3385	5.9060	220	0.3887
0.3385	6.1745	230	0.3960
0.2922	6.4430	240	0.3805
0.2922	6.7114	250	0.3767
0.253	6.9799	260	0.3760
0.253	7.2483	270	0.3811
0.2518	7.5168	280	0.3757
0.2518	7.7852	290	0.3635
0.2191	8.0537	300	0.3677
0.2191	8.3221	310	0.3711
0.223	8.5906	320	0.3700
0.223	8.8591	330	0.3685
0.1998	9.1275	340	0.3636
0.1998	9.3960	350	0.3622
0.2208	9.6644	360	0.3632
0.2208	9.9329	370	0.3637

Framework versions

—PEFT 0.10.0
—Transformers 4.40.2
—Pytorch 2.2.1+cu121
—Datasets 2.19.1
—Tokenizers 0.19.1