CodeLlama-7b-Instruct-hf_En__size_52_epochs_10_2024-06-21_06-58-10_3556411

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: 1.6041
• Accuracy: 0.054
• Chrf: 0.699
• Bleu: 0.622
• Sacrebleu: 0.6
• Rouge1: 0.691
• Rouge2: 0.483
• Rougel: 0.637
• Rougelsum: 0.682
• Meteor: 0.56

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: 52
• training_steps: 520

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.1805	4.0	52	1.4057	0.058	0.674	0.548	0.5	0.651	0.411	0.596	0.643	0.529
0.9191	8.0	104	1.9644	0.05	0.619	0.49	0.5	0.593	0.36	0.55	0.588	0.503
0.3517	12.0	156	1.8542	0.052	0.628	0.526	0.5	0.63	0.407	0.582	0.625	0.522
0.4017	16.0	208	2.2165	0.057	0.565	0.428	0.4	0.529	0.276	0.474	0.523	0.462
0.324	20.0	260	1.8054	0.055	0.648	0.551	0.6	0.631	0.415	0.586	0.623	0.527
0.5071	24.0	312	1.7591	0.058	0.671	0.562	0.6	0.651	0.435	0.599	0.644	0.531
0.1758	28.0	364	1.6743	0.054	0.683	0.585	0.6	0.671	0.464	0.62	0.663	0.553
0.4696	32.0	416	1.6739	0.055	0.679	0.592	0.6	0.656	0.44	0.606	0.645	0.536
0.1516	36.0	468	1.6355	0.054	0.689	0.611	0.6	0.679	0.473	0.627	0.669	0.554
0.3236	40.0	520	1.6041	0.054	0.699	0.622	0.6	0.691	0.483	0.637	0.682	0.56

Framework versions

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