lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_meta-llama_Llama-2-7b-hf_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_reciteonly_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 1.7481
• Accuracy: 0.6453

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: 20.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.4597	1.0	187	1.2719	0.6619
1.2157	2.0	375	1.2062	0.6670
1.1861	3.0	562	1.1983	0.6675
1.1445	4.0	750	1.1977	0.6672
1.1	5.0	937	1.2079	0.6665
1.0439	6.0	1125	1.2238	0.6650
0.9888	7.0	1312	1.2457	0.6638
0.9364	8.0	1500	1.2816	0.6616
0.8889	9.0	1687	1.3036	0.6606
0.8373	10.0	1875	1.3337	0.6587
0.7949	11.0	2062	1.3678	0.6575
0.7539	12.0	2250	1.3984	0.6554
0.7133	13.0	2437	1.4471	0.6538
0.6704	14.0	2625	1.4830	0.6525
0.6436	15.0	2812	1.5243	0.6508
0.6073	16.0	3000	1.5641	0.6500
0.5567	17.0	3187	1.6059	0.6487
0.5204	18.0	3375	1.6426	0.6473
0.5019	19.0	3562	1.6984	0.6460
0.4676	19.95	3740	1.7481	0.6453

Framework versions

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