lmind_hotpot_train8000_eval7405_v1_recite_qa_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_hotpot_train8000_eval7405_v1_recite_qa dataset. It achieves the following results on the evaluation set:

• Loss: 0.4673
• Accuracy: 0.7786

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
1.0721	1.0	1089	1.0138	0.7178
1.0225	2.0	2178	0.9785	0.7211
0.9751	3.0	3267	0.9417	0.7243
0.909	4.0	4357	0.9046	0.7279
0.85	5.0	5446	0.8675	0.7314
0.782	6.0	6535	0.8283	0.7349
0.7153	7.0	7624	0.7927	0.7383
0.6625	8.0	8714	0.7558	0.7425
0.6041	9.0	9803	0.7214	0.7462
0.5372	10.0	10892	0.6826	0.7502
0.4811	11.0	11981	0.6502	0.7539
0.443	12.0	13071	0.6224	0.7571
0.3957	13.0	14160	0.5878	0.7611
0.3511	14.0	15249	0.5610	0.7642
0.3205	15.0	16338	0.5359	0.7671
0.2947	16.0	17428	0.5196	0.7701
0.2657	17.0	18517	0.4951	0.7730
0.2446	18.0	19606	0.4904	0.7743
0.2248	19.0	20695	0.4748	0.7769
0.2044	20.0	21780	0.4673	0.7786

Framework versions

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