lmind_nq_train6000_eval6489_v1_docidx_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_docidx_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 4.4400
• Accuracy: 0.4387

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
1.3957	1.0	341	3.3998	0.4541
1.3756	2.0	683	3.4568	0.4546
1.3109	3.0	1024	3.5541	0.4578
1.2488	4.0	1366	3.6057	0.4573
1.1856	5.0	1707	3.7215	0.4557
1.1284	6.0	2049	3.7284	0.4545
1.0567	7.0	2390	3.8020	0.4533
0.978	8.0	2732	3.8535	0.4524
0.9007	9.0	3073	3.9364	0.4516
0.833	10.0	3415	3.9463	0.4499
0.7455	11.0	3756	4.0375	0.4488
0.6909	12.0	4098	4.1021	0.4471
0.6243	13.0	4439	4.1491	0.4457
0.5672	14.0	4781	4.2086	0.4441
0.5096	15.0	5122	4.2696	0.4443
0.4532	16.0	5464	4.2835	0.4422
0.4201	17.0	5805	4.3720	0.4411
0.3642	18.0	6147	4.3791	0.4412
0.3222	19.0	6488	4.4365	0.4393
0.2966	19.97	6820	4.4400	0.4387

Framework versions

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