OpenCodeInterpreter-DS-6.7B_Fi__translations_size_104_epochs_10_2024-06-21_23-19-55_3557637

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

• Loss: 3.0113
• Accuracy: 0.03
• Chrf: 0.519
• Bleu: 0.428
• Sacrebleu: 0.4
• Rouge1: 0.51
• Rouge2: 0.267
• Rougel: 0.473
• Rougelsum: 0.504
• Meteor: 0.393

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
0.1566	4.0	104	1.1892	0.011	0.724	0.584	0.6	0.662	0.439	0.599	0.654	0.493
0.1145	8.0	208	1.3290	0.015	0.697	0.556	0.6	0.643	0.393	0.569	0.635	0.515
0.143	12.0	312	1.4376	0.016	0.686	0.577	0.6	0.673	0.453	0.609	0.669	0.516
0.6387	16.0	416	1.6801	0.018	0.723	0.599	0.6	0.69	0.49	0.634	0.684	0.508
0.1887	20.0	520	2.0277	0.023	0.674	0.551	0.6	0.663	0.441	0.595	0.66	0.48
0.2298	24.0	624	2.3972	0.028	0.621	0.512	0.5	0.621	0.397	0.562	0.617	0.452
1.0534	28.0	728	2.7032	0.024	0.578	0.479	0.5	0.577	0.344	0.53	0.565	0.427
0.2948	32.0	832	2.9091	0.031	0.545	0.452	0.5	0.54	0.302	0.498	0.535	0.415
0.3843	36.0	936	3.0397	0.03	0.498	0.407	0.4	0.498	0.248	0.456	0.493	0.388
0.3182	40.0	1040	3.0113	0.03	0.519	0.428	0.4	0.51	0.267	0.473	0.504	0.393

Framework versions

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