lmind_nq_train6000_eval6489_v1_doc_qa_v3_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_nq_train6000_eval6489_v1_doc_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 10.0702
• Accuracy: 0.1692

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
1.3822	1.0	529	1.2977	0.6172
1.2744	2.0	1058	1.3745	0.6032
1.1768	3.0	1587	1.3319	0.6157
0.9247	4.0	2116	1.4367	0.6102
1.1836	5.0	2645	1.9168	0.5569
2.035	6.0	3174	2.0794	0.5377
3.7483	7.0	3703	2.6723	0.4881
7.127	8.0	4232	7.0410	0.1922
7.5321	9.0	4761	6.6488	0.1941
7.3806	10.0	5290	6.8427	0.2197
7.8159	11.0	5819	6.8836	0.2197
7.975	12.0	6348	6.8763	0.2197
7.9902	13.0	6877	6.8726	0.2197
7.8585	14.0	7406	6.8236	0.2195
7.3449	15.0	7935	7.1997	0.1922
7.3133	16.0	8464	6.7455	0.1869
7.305	17.0	8993	6.7454	0.1869
7.7463	18.0	9522	8.8319	0.1870
9.9696	19.0	10051	10.0702	0.1692
9.9845	20.0	10580	10.0702	0.1692

Framework versions

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