CodeLlama-13b-Instruct-hf_Fi__components_size_252_epochs_10_2024-06-21_09-35-18_3556546

This model is a fine-tuned version of codellama/CodeLlama-13b-Instruct-hf on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7475
• Accuracy: 0.464
• Chrf: 0.369
• Bleu: 0.302
• Sacrebleu: 0.3
• Rouge1: 0.478
• Rouge2: 0.294
• Rougel: 0.457
• Rougelsum: 0.472
• Meteor: 0.575

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: 252
• training_steps: 2520

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.05	4.0	252	2.7480	0.47	0.15	0.149	0.1	0.278	0.09	0.264	0.275	0.338
0.0657	8.0	504	2.3686	0.46	0.23	0.189	0.2	0.426	0.221	0.395	0.424	0.472
0.0785	12.0	756	2.5520	0.467	0.2	0.166	0.2	0.412	0.243	0.373	0.396	0.445
1.0703	16.0	1008	2.2775	0.46	0.27	0.205	0.2	0.452	0.269	0.426	0.451	0.494
0.1709	20.0	1260	2.2827	0.482	0.276	0.224	0.2	0.405	0.252	0.399	0.403	0.476
0.3803	24.0	1512	2.0205	0.463	0.28	0.241	0.2	0.424	0.277	0.411	0.417	0.522
0.0239	28.0	1764	1.8601	0.462	0.343	0.288	0.3	0.47	0.293	0.45	0.47	0.554
0.1383	32.0	2016	1.8422	0.464	0.358	0.301	0.3	0.487	0.306	0.459	0.486	0.568
0.0313	36.0	2268	1.7820	0.461	0.363	0.295	0.3	0.483	0.304	0.467	0.482	0.582
0.0352	40.0	2520	1.7475	0.464	0.369	0.302	0.3	0.478	0.294	0.457	0.472	0.575

Framework versions

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