CodeLlama-7b-Instruct-hf_Fi__translations_size_104_epochs_10_2024-06-22_03-11-17_3557642

This model is a fine-tuned version of codellama/CodeLlama-7b-Instruct-hf on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.3204
• Accuracy: 0.035
• Chrf: 0.736
• Bleu: 0.656
• Sacrebleu: 0.7
• Rouge1: 0.689
• Rouge2: 0.506
• Rougel: 0.657
• Rougelsum: 0.684
• Meteor: 0.613

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
• train_batch_size: 1
• eval_batch_size: 1
• seed: 3407
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 4
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 104
• training_steps: 1040

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.2911	4.0	104	1.6937	0.042	0.623	0.509	0.5	0.586	0.333	0.536	0.584	0.513
0.1827	8.0	208	1.9669	0.04	0.584	0.47	0.5	0.532	0.309	0.495	0.528	0.48
0.1801	12.0	312	1.7206	0.039	0.621	0.53	0.5	0.582	0.353	0.55	0.579	0.538
0.5158	16.0	416	1.5969	0.038	0.654	0.536	0.5	0.602	0.368	0.565	0.599	0.52
0.1586	20.0	520	1.6966	0.037	0.632	0.522	0.5	0.576	0.357	0.544	0.571	0.513
0.1584	24.0	624	1.5539	0.037	0.694	0.597	0.6	0.651	0.447	0.61	0.643	0.564
0.6664	28.0	728	1.4467	0.035	0.701	0.611	0.6	0.655	0.452	0.616	0.651	0.574
0.171	32.0	832	1.6334	0.033	0.68	0.568	0.6	0.633	0.417	0.59	0.627	0.526
0.227	36.0	936	1.3442	0.035	0.727	0.64	0.6	0.682	0.491	0.645	0.679	0.577
0.1831	40.0	1040	1.3204	0.035	0.736	0.656	0.7	0.689	0.506	0.657	0.684	0.613

Framework versions

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