lmind_nq_train6000_eval6489_v1_doc_qa_v3_3e-5_lora2

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the tyzhu/lmind_nq_train6000_eval6489_v1_doc_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0598
• Accuracy: 0.5423

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	Accuracy	Validation Loss
1.3948	1.0	529	0.6132	1.3087
1.3789	2.0	1058	0.6146	1.2897
1.3259	3.0	1587	0.6179	1.2849
1.2853	4.0	2116	0.6159	1.3169
1.2556	5.0	2645	0.6132	1.3532
1.1972	6.0	3174	0.6126	1.4135
1.1839	7.0	3703	0.6081	1.5007
1.1334	8.0	4232	0.6074	1.5242
1.0966	9.0	4761	0.5803	1.6107
1.0485	10.0	5290	0.6049	1.6749
1.021	11.0	5819	0.6015	1.7324
0.9918	12.0	6348	0.6007	1.7632
0.947	13.0	6877	0.6011	1.8303
0.9376	14.0	7406	0.5991	1.8873
0.898	15.0	7935	0.5976	1.9688
0.8559	16.0	8464	0.5988	1.9724
0.8348	17.0	8993	0.5714	1.9815
0.8106	18.0	9522	0.598	2.0386
0.7848	19.0	10051	0.5964	2.0627
0.745	20.0	10580	0.5966	2.0825
0.7208	21.0	11109	0.5959	2.0959
0.6842	22.0	11638	0.5968	2.1534
0.6661	23.0	12167	0.5975	2.1792
0.6193	24.0	12696	0.5967	2.1530
0.6064	25.0	13225	0.5958	2.1720
0.5776	26.0	13754	0.5966	2.2162
0.5492	27.0	14283	0.5862	2.2382
0.5256	28.0	14812	0.5963	2.2273
0.5128	29.0	15341	0.5948	2.2448
0.4846	30.0	15870	0.5846	2.2697
0.4623	31.0	16399	0.5968	2.2425
0.4468	32.0	16928	0.5936	2.2654
0.4714	33.0	17457	0.5957	2.1317
0.8308	34.0	17986	0.5973	1.9392
0.6478	35.0	18515	0.5979	2.0346
0.612	36.0	19044	0.5978	2.0882
0.5928	37.0	19573	0.5970	2.1420
0.5698	38.0	20102	0.5966	2.1569
0.5444	39.0	20631	0.5956	2.1954
0.5404	40.0	21160	0.5942	2.1724
0.5124	41.0	21689	0.5939	2.2020
0.5342	42.0	22218	0.5938	2.1955
0.5385	43.0	22747	0.5946	2.1431
0.5673	44.0	23276	0.5948	2.1269
0.7034	45.0	23805	0.5917	2.0601
1.0751	46.0	24334	0.5861	1.8910
1.9072	47.0	24863	0.5518	2.1388
5.2339	48.0	25392	0.3825	4.4877
2.573	49.0	25921	0.5283	2.2255
2.1439	50.0	26450	0.5423	2.0598

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.14.1