lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_meta-llama_Llama-2-7b-hf_5e-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: 2.1485
• Accuracy: 0.6417

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: 5e-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
1.3988	1.0	187	1.2107	0.6661
1.1977	2.0	375	1.1987	0.6676
1.1458	3.0	562	1.1964	0.6676
1.0707	4.0	750	1.2088	0.6666
1.0066	5.0	937	1.2342	0.6652
0.9353	6.0	1125	1.2646	0.6628
0.8629	7.0	1312	1.2991	0.6617
0.7933	8.0	1500	1.3467	0.6596
0.7185	9.0	1687	1.3917	0.6575
0.6489	10.0	1875	1.4609	0.6550
0.586	11.0	2062	1.5205	0.6531
0.5267	12.0	2250	1.5616	0.6518
0.4702	13.0	2437	1.6518	0.6499
0.42	14.0	2625	1.7193	0.6481
0.3811	15.0	2812	1.8016	0.6459
0.3373	16.0	3000	1.8556	0.6465
0.284	17.0	3187	1.9407	0.6443
0.25	18.0	3375	1.9760	0.6436
0.2274	19.0	3562	2.1003	0.6424
0.2033	19.95	3740	2.1485	0.6417

Framework versions

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