CodeLlama-7b-Instruct-hf_En__components_size_252_epochs_10_2024-06-21_16-51-46_3556559

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

• Loss: 1.9251
• Accuracy: 0.496
• Chrf: 0.315
• Bleu: 0.248
• Sacrebleu: 0.2
• Rouge1: 0.458
• Rouge2: 0.26
• Rougel: 0.435
• Rougelsum: 0.456
• Meteor: 0.521

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
0.0421	4.0	252	3.2663	0.495	0.084	0.035	0.0	0.203	0.043	0.18	0.192	0.201
0.0562	8.0	504	2.4211	0.495	0.177	0.133	0.1	0.352	0.155	0.332	0.352	0.457
0.08	12.0	756	2.7082	0.494	0.184	0.08	0.1	0.242	0.082	0.23	0.241	0.33
0.9994	16.0	1008	2.4576	0.496	0.216	0.116	0.1	0.4	0.217	0.381	0.399	0.398
0.1476	20.0	1260	2.7555	0.497	0.159	0.051	0.1	0.274	0.064	0.259	0.262	0.196
0.3371	24.0	1512	2.2083	0.491	0.196	0.128	0.1	0.422	0.217	0.396	0.414	0.384
0.0187	28.0	1764	2.0562	0.476	0.268	0.19	0.2	0.435	0.238	0.413	0.431	0.462
0.1243	32.0	2016	2.0119	0.497	0.291	0.238	0.2	0.454	0.248	0.424	0.452	0.504
0.0295	36.0	2268	1.9499	0.479	0.309	0.244	0.2	0.443	0.253	0.43	0.441	0.527
0.027	40.0	2520	1.9251	0.496	0.315	0.248	0.2	0.458	0.26	0.435	0.456	0.521

Framework versions

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