Modelpublic

vdavidr/CodeLlama-13b-Instruct-hf_En__translations_size_104_epochs_10_2024-06-22_07-00-27_3558000

sourceHugging Facellama2updated 2y agoView on Hugging Face

0likes

Model Card

CodeLlama-13b-Instruct-hfEntranslationssize104epochs102024-06-2207-00-273558000

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: 1.7291
—Accuracy: 0.053
—Chrf: 0.654
—Bleu: 0.554
—Sacrebleu: 0.6
—Rouge1: 0.61
—Rouge2: 0.367
—Rougel: 0.556
—Rougelsum: 0.603
—Meteor: 0.542

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: 0.001
—trainbatchsize: 1
—evalbatchsize: 1
—seed: 3407
—distributed_type: multi-GPU
—num_devices: 4
—totaltrainbatch_size: 4
—totalevalbatch_size: 4
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06
—lrschedulertype: linear
—lrschedulerwarmup_steps: 104
—training_steps: 1040

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.2903	4.0	104	2.0336	0.052	0.541	0.417	0.4	0.524	0.268	0.48	0.519	0.477
0.2305	8.0	208	2.2999	0.05	0.513	0.412	0.4	0.495	0.26	0.454	0.489	0.453
0.1581	12.0	312	2.0944	0.054	0.566	0.475	0.5	0.548	0.313	0.502	0.543	0.52
0.6402	16.0	416	1.9048	0.05	0.601	0.482	0.5	0.568	0.32	0.521	0.562	0.544
0.1487	20.0	520	2.0162	0.054	0.591	0.487	0.5	0.567	0.324	0.516	0.556	0.472
0.1857	24.0	624	2.0241	0.055	0.572	0.481	0.5	0.545	0.304	0.499	0.542	0.509
0.718	28.0	728	1.7555	0.054	0.63	0.529	0.5	0.595	0.345	0.549	0.588	0.532
0.1329	32.0	832	1.7785	0.053	0.651	0.554	0.6	0.619	0.379	0.567	0.614	0.536
0.2488	36.0	936	1.7203	0.052	0.657	0.556	0.6	0.608	0.371	0.558	0.599	0.549
0.1337	40.0	1040	1.7291	0.053	0.654	0.554	0.6	0.61	0.367	0.556	0.603	0.542

Framework versions

—Transformers 4.37.0
—Pytorch 2.2.1+cu121
—Datasets 2.20.0
—Tokenizers 0.15.2