lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_meta-llama_Llama-2-7b-hf_5e-4_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: 7.3304
• Accuracy: 0.2409

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: 0.0005
• 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
1.2462	1.0	187	1.1816	0.6696
0.9978	2.0	375	1.1967	0.6695
0.6981	3.0	562	1.2826	0.6646
0.437	4.0	750	1.4432	0.6590
0.2655	5.0	937	1.6258	0.6546
0.1738	6.0	1125	1.7712	0.6520
0.1308	7.0	1312	1.8679	0.6522
0.1073	8.0	1500	2.0001	0.6507
0.0895	9.0	1687	2.0216	0.6508
0.0873	10.0	1875	2.0699	0.6486
0.1112	11.0	2062	2.0632	0.6489
0.144	12.0	2250	2.0565	0.6458
7.3681	13.0	2437	7.3700	0.2426
7.3792	14.0	2625	7.3533	0.2428
7.3938	15.0	2812	7.3602	0.2410
7.4413	16.0	3000	7.4299	0.2411
7.4344	17.0	3187	7.3487	0.2440
7.3659	18.0	3375	7.3300	0.2433
7.3585	19.0	3562	7.3320	0.2414
7.349	19.95	3740	7.3304	0.2409

Framework versions

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