OpenCodeInterpreter-DS-6.7B_En__size_52_epochs_10_2024-06-21_05-24-05_3556406

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: 1.3571
• Accuracy: 0.039
• Chrf: 0.755
• Bleu: 0.641
• Sacrebleu: 0.6
• Rouge1: 0.733
• Rouge2: 0.524
• Rougel: 0.664
• Rougelsum: 0.721
• Meteor: 0.587

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
0.1314	4.0	52	1.2192	0.03	0.713	0.585	0.6	0.655	0.426	0.587	0.643	0.523
0.5793	8.0	104	1.2889	0.032	0.721	0.577	0.6	0.668	0.435	0.595	0.653	0.507
0.1913	12.0	156	1.1944	0.031	0.715	0.578	0.6	0.662	0.437	0.593	0.649	0.519
0.1824	16.0	208	1.2273	0.038	0.725	0.588	0.6	0.684	0.447	0.609	0.665	0.534
0.1701	20.0	260	1.2799	0.038	0.727	0.587	0.6	0.688	0.456	0.617	0.67	0.505
0.4458	24.0	312	1.3126	0.037	0.73	0.6	0.6	0.695	0.461	0.618	0.679	0.52
0.1128	28.0	364	1.2848	0.038	0.752	0.628	0.6	0.716	0.504	0.647	0.703	0.54
0.4474	32.0	416	1.3747	0.039	0.748	0.619	0.6	0.715	0.49	0.643	0.703	0.543
0.1126	36.0	468	1.3679	0.039	0.752	0.63	0.6	0.725	0.506	0.651	0.712	0.549
0.2454	40.0	520	1.3571	0.039	0.755	0.641	0.6	0.733	0.524	0.664	0.721	0.587

Framework versions

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