lmind_nq_train6000_eval6489_v1_recite_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_recite_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5108
• Accuracy: 0.7842

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
1.3072	1.0	529	1.1908	0.6659
1.2561	2.0	1058	1.1422	0.6728
1.1672	3.0	1587	1.0947	0.6793
1.0827	4.0	2116	1.0515	0.6857
0.9944	5.0	2645	1.0024	0.6940
0.8926	6.0	3174	0.9521	0.7009
0.8135	7.0	3703	0.8976	0.7097
0.7146	8.0	4232	0.8515	0.7171
0.6265	9.0	4761	0.8022	0.7259
0.5522	10.0	5290	0.7628	0.7327
0.4864	11.0	5819	0.7115	0.7419
0.4265	12.0	6348	0.6697	0.7488
0.3671	13.0	6877	0.6299	0.7560
0.331	14.0	7406	0.6025	0.7616
0.2923	15.0	7935	0.5802	0.7667
0.2525	16.0	8464	0.5576	0.7711
0.2353	17.0	8993	0.5441	0.7753
0.2121	18.0	9522	0.5286	0.7796
0.1936	19.0	10051	0.5184	0.7825
0.1784	20.0	10580	0.5108	0.7842

Framework versions

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