lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_3e-5_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 3.5443
• Accuracy: 0.5730

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: 3e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 50.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.8545	0.9973	187	1.7123	0.6037
1.7063	2.0	375	1.6945	0.6059
1.6702	2.9973	562	1.6851	0.6066
1.6356	4.0	750	1.6826	0.6075
1.5775	4.9973	937	1.6911	0.6071
1.529	6.0	1125	1.7067	0.6069
1.457	6.9973	1312	1.7279	0.6056
1.3907	8.0	1500	1.7512	0.6046
1.3309	8.9973	1687	1.7774	0.6025
1.2841	10.0	1875	1.8043	0.6013
1.2308	10.9973	2062	1.8528	0.6001
1.1722	12.0	2250	1.8851	0.5988
1.1354	12.9973	2437	1.9114	0.5980
1.0793	14.0	2625	1.9585	0.5961
1.037	14.9973	2812	1.9967	0.5948
0.9901	16.0	3000	2.0336	0.5934
0.9316	16.9973	3187	2.0880	0.5914
0.8802	18.0	3375	2.1440	0.5901
0.8382	18.9973	3562	2.1715	0.5893
0.7962	20.0	3750	2.2237	0.5879
0.7553	20.9973	3937	2.2957	0.5861
0.7238	22.0	4125	2.3312	0.5851
0.676	22.9973	4312	2.4043	0.5832
0.644	24.0	4500	2.4440	0.5824
0.5939	24.9973	4687	2.5127	0.5818
0.5551	26.0	4875	2.5390	0.5810
0.5163	26.9973	5062	2.5809	0.5798
0.4892	28.0	5250	2.6670	0.5789
0.4669	28.9973	5437	2.6695	0.5786
0.4353	30.0	5625	2.7646	0.5787
0.4104	30.9973	5812	2.8291	0.5775
0.3885	32.0	6000	2.8933	0.5764
0.342	32.9973	6187	2.9434	0.5756
0.3213	34.0	6375	2.9346	0.5756
0.3065	34.9973	6562	3.0082	0.5758
0.2842	36.0	6750	3.0947	0.5739
0.2695	36.9973	6937	3.0905	0.5752
0.2541	38.0	7125	3.1831	0.5738
0.2411	38.9973	7312	3.2135	0.5740
0.228	40.0	7500	3.2505	0.5739
0.2067	40.9973	7687	3.2867	0.5743
0.1952	42.0	7875	3.3047	0.5751
0.1886	42.9973	8062	3.3528	0.5742
0.1828	44.0	8250	3.4431	0.5730
0.1743	44.9973	8437	3.4166	0.5727
0.1691	46.0	8625	3.4326	0.5739
0.1633	46.9973	8812	3.4555	0.5728
0.156	48.0	9000	3.4876	0.5729
0.1441	48.9973	9187	3.5368	0.5727
0.1401	49.8667	9350	3.5443	0.5730

Framework versions

• PEFT 0.5.0
• Transformers 4.41.1
• Pytorch 2.1.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1