lmind_hotpot_train8000_eval7405_v1_docidx_meta-llama_Llama-2-7b-hf_3e-5_lora2

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the tyzhu/lmind_hotpot_train8000_eval7405_v1_docidx dataset. It achieves the following results on the evaluation set:

• Loss: 1.0143
• Accuracy: 0.7798

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: 3e-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
1.1149	1.0	839	1.3354	0.7522
1.0834	2.0	1678	1.3234	0.7536
1.0714	3.0	2517	1.2758	0.7554
1.0547	4.0	3357	1.2809	0.7570
1.0218	5.0	4196	1.2560	0.7587
0.9748	6.0	5035	1.2244	0.7605
0.9439	7.0	5874	1.1968	0.7621
0.9155	8.0	6714	1.1888	0.7634
0.8767	9.0	7553	1.1624	0.7651
0.848	10.0	8392	1.1410	0.7664
0.8227	11.0	9231	1.1288	0.7678
0.7874	12.0	10071	1.1165	0.7694
0.7469	13.0	10910	1.1008	0.7703
0.7256	14.0	11749	1.0892	0.7721
0.701	15.0	12588	1.0651	0.7732
0.6651	16.0	13428	1.0723	0.7746
0.6534	17.0	14267	1.0388	0.7759
0.6092	18.0	15106	1.0244	0.7774
0.5946	19.0	15945	1.0292	0.7784
0.5526	19.99	16780	1.0143	0.7798

Framework versions

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