lmind_hotpot_train8000_eval7405_v1_reciteonly_qa_meta-llama_Llama-2-7b-hf_3e-4_lora2

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the tyzhu/lmind_hotpot_train8000_eval7405_v1_reciteonly_qa dataset. It achieves the following results on the evaluation set:

• Loss: 2.0861
• Accuracy: 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: 0.0003
• 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
1.0107	1.0	250	1.0139	0.7185
0.9083	2.0	500	1.0163	0.7184
0.7726	3.0	750	1.0599	0.7162
0.6247	4.0	1000	1.1299	0.7134
0.4763	5.0	1250	1.2265	0.7103
0.3518	6.0	1500	1.3529	0.7074
0.2427	7.0	1750	1.4919	0.7054
0.1766	8.0	2000	1.6280	0.7034
0.1259	9.0	2250	1.7092	0.7029
0.1058	10.0	2500	1.8115	0.7023
0.0815	11.0	2750	1.9252	0.7014
0.0749	12.0	3000	1.9652	0.7015
0.0674	13.0	3250	2.0071	0.7007
0.0678	14.0	3500	2.0076	0.7009
0.0642	15.0	3750	2.0298	0.7013
0.0656	16.0	4000	2.0544	0.7006
0.0617	17.0	4250	2.0874	0.7010
0.0635	18.0	4500	2.0979	0.7002
0.0602	19.0	4750	2.0700	0.7004
0.123	20.0	5000	2.0861	0.6999

Framework versions

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