README.md

metadata

library_name: transformers
license: llama2
base_model: llava-hf/llava-1.5-7b-hf
tags:
  - trl
  - sft
  - generated_from_trainer
metrics:
  - bleu
  - rouge
model-index:
  - name: sft-llava-1.5-7b_lora
    results: []

sft-llava-1.5-7b_lora

This model is a fine-tuned version of llava-hf/llava-1.5-7b-hf on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.9404
• Bleu: 0.1802
• Rouge1: 0.4861
• Rouge2: 0.1709
• Rougel: 0.3580
• Bertscore Precision: 0.6578
• Bertscore Recall: 0.7479
• Bertscore F1: 0.6999

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 5.0

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Rouge1	Rouge2	Rougel	Bertscore Precision	Bertscore Recall	Bertscore F1
5.7514	0.3101	200	5.6831	0.0772	0.2028	0.0717	0.1778	0.6381	0.7437	0.6869
2.9737	0.6202	400	2.9242	0.1580	0.4319	0.1445	0.3306	0.6578	0.7479	0.6999
2.6756	0.9302	600	2.6594	0.1839	0.4859	0.1759	0.3680	0.6381	0.7437	0.6869
2.18	1.2403	800	2.5783	0.1754	0.4864	0.1754	0.3775	0.6578	0.7479	0.6999
2.0957	1.5504	1000	2.5019	0.1849	0.4877	0.1850	0.3801	0.6578	0.7479	0.6999
2.0109	1.8605	1200	2.4393	0.1879	0.4911	0.1840	0.3859	0.6578	0.7479	0.6999
0.7656	2.1705	1400	2.9613	0.1808	0.4810	0.1719	0.3644	0.6578	0.7479	0.6999
0.7271	2.4806	1600	3.0544	0.1817	0.4795	0.1695	0.3629	0.6578	0.7479	0.6999
0.6746	2.7907	1800	3.0377	0.1754	0.4765	0.1639	0.3508	0.6578	0.7479	0.6999
0.1183	3.1008	2000	3.6408	0.1801	0.4821	0.1710	0.3636	0.6578	0.7479	0.6999
0.1123	3.4109	2200	3.6913	0.1765	0.4903	0.1712	0.3629	0.6578	0.7479	0.6999
0.1051	3.7209	2400	3.7181	0.1766	0.4884	0.1701	0.3618	0.6578	0.7479	0.6999
0.046	4.0310	2600	3.7719	0.1781	0.4849	0.1711	0.3598	0.6578	0.7479	0.6999
0.0444	4.3411	2800	3.9170	0.1801	0.4852	0.1719	0.3595	0.6578	0.7479	0.6999
0.0452	4.6512	3000	3.9377	0.1808	0.4872	0.1714	0.3604	0.6578	0.7479	0.6999
0.0449	4.9612	3200	3.9404	0.1802	0.4861	0.1709	0.3580	0.6578	0.7479	0.6999

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.0a0+81ea7a4
• Datasets 3.0.1
• Tokenizers 0.20.1