CodeQwen1.5-7B-Chat_components_FI_dataset_size_52_epochs_10_2024-06-13_03-26-06_22016831

This model is a fine-tuned version of Qwen/CodeQwen1.5-7B-Chat on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2276
• Accuracy: 0.477
• Chrf: 0.858
• Bleu: 0.793
• Sacrebleu: 0.8
• Rouge1: 0.842
• Rouge2: 0.743
• Rougel: 0.827
• Rougelsum: 0.838
• Meteor: 0.844

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 52
• training_steps: 520
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.0684	0.2063	52	0.4553	0.478	0.777	0.686	0.7	0.716	0.555	0.679	0.709	0.731
0.0098	0.4127	104	0.3791	0.478	0.799	0.72	0.7	0.778	0.641	0.746	0.773	0.777
0.0066	0.6190	156	0.3244	0.477	0.818	0.743	0.7	0.809	0.689	0.786	0.804	0.797
0.0009	0.8254	208	0.2773	0.477	0.839	0.769	0.8	0.82	0.705	0.801	0.815	0.823
0.0104	1.0317	260	0.2711	0.477	0.845	0.777	0.8	0.83	0.718	0.81	0.825	0.827
0.0192	1.2381	312	0.2536	0.477	0.844	0.777	0.8	0.827	0.719	0.812	0.824	0.832
0.0011	1.4444	364	0.2491	0.477	0.846	0.78	0.8	0.835	0.729	0.82	0.831	0.82
0.0021	1.6508	416	0.2348	0.477	0.854	0.786	0.8	0.834	0.723	0.817	0.829	0.838
0.0007	1.8571	468	0.2331	0.477	0.853	0.786	0.8	0.834	0.724	0.818	0.83	0.836
0.0862	2.0635	520	0.2276	0.477	0.858	0.793	0.8	0.842	0.743	0.827	0.838	0.844

Framework versions

• PEFT 0.7.1
• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1