Artigenz-Coder-DS-6.7B_En__size_52_epochs_10_2024-06-21_05-38-04_3556407

This model is a fine-tuned version of Artigenz/Artigenz-Coder-DS-6.7B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5021
• Accuracy: 0.035
• Chrf: 0.723
• Bleu: 0.604
• Sacrebleu: 0.6
• Rouge1: 0.692
• Rouge2: 0.471
• Rougel: 0.627
• Rougelsum: 0.677
• Meteor: 0.491

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
0.1281	4.0	52	1.2181	0.029	0.733	0.612	0.6	0.674	0.454	0.608	0.665	0.546
0.5716	8.0	104	1.2202	0.032	0.714	0.602	0.6	0.676	0.461	0.612	0.667	0.503
0.1833	12.0	156	1.1088	0.036	0.722	0.613	0.6	0.689	0.468	0.619	0.675	0.552
0.1646	16.0	208	1.1923	0.034	0.729	0.614	0.6	0.699	0.47	0.623	0.684	0.545
0.171	20.0	260	1.4073	0.034	0.7	0.591	0.6	0.662	0.447	0.599	0.649	0.471
0.4405	24.0	312	1.4275	0.035	0.704	0.588	0.6	0.663	0.447	0.601	0.651	0.482
0.1094	28.0	364	1.3566	0.035	0.727	0.619	0.6	0.692	0.481	0.628	0.678	0.52
0.5148	32.0	416	1.3953	0.034	0.728	0.622	0.6	0.697	0.481	0.631	0.683	0.527
0.1122	36.0	468	1.4576	0.035	0.72	0.607	0.6	0.693	0.466	0.622	0.68	0.512
0.246	40.0	520	1.5021	0.035	0.723	0.604	0.6	0.692	0.471	0.627	0.677	0.491

Framework versions

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