Artigenz-Coder-DS-6.7B_Fi__translations_size_104_epochs_10_2024-06-21_23-22-47_3557638

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: 2.9722
• Accuracy: 0.038
• Chrf: 0.522
• Bleu: 0.437
• Sacrebleu: 0.4
• Rouge1: 0.515
• Rouge2: 0.259
• Rougel: 0.472
• Rougelsum: 0.507
• Meteor: 0.41

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
0.1488	4.0	104	1.1476	0.024	0.727	0.608	0.6	0.69	0.485	0.629	0.684	0.537
0.0978	8.0	208	1.1940	0.014	0.721	0.6	0.6	0.675	0.466	0.607	0.668	0.53
0.1307	12.0	312	1.3964	0.017	0.708	0.596	0.6	0.677	0.464	0.612	0.668	0.532
0.5772	16.0	416	1.6675	0.02	0.71	0.62	0.6	0.697	0.5	0.635	0.691	0.529
0.1787	20.0	520	2.0574	0.043	0.664	0.556	0.6	0.65	0.429	0.586	0.642	0.493
0.2195	24.0	624	2.4599	0.038	0.624	0.533	0.5	0.623	0.395	0.57	0.616	0.461
1.0463	28.0	728	2.7031	0.029	0.564	0.468	0.5	0.569	0.33	0.513	0.559	0.436
0.2775	32.0	832	2.8662	0.038	0.551	0.469	0.5	0.56	0.303	0.504	0.549	0.422
0.3616	36.0	936	2.9521	0.037	0.535	0.446	0.4	0.524	0.27	0.479	0.516	0.415
0.3042	40.0	1040	2.9722	0.038	0.522	0.437	0.4	0.515	0.259	0.472	0.507	0.41

Framework versions

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