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

• Loss: 2.0825
• Accuracy: 0.5966

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
1.3948	1.0	529	1.3087	0.6132
1.3789	2.0	1058	1.2897	0.6146
1.3259	3.0	1587	1.2849	0.6179
1.2853	4.0	2116	1.3169	0.6159
1.2556	5.0	2645	1.3532	0.6132
1.1972	6.0	3174	1.4135	0.6126
1.1839	7.0	3703	1.5007	0.6081
1.1334	8.0	4232	1.5242	0.6074
1.0966	9.0	4761	1.6107	0.5803
1.0485	10.0	5290	1.6749	0.6049
1.021	11.0	5819	1.7324	0.6015
0.9918	12.0	6348	1.7632	0.6007
0.947	13.0	6877	1.8303	0.6011
0.9376	14.0	7406	1.8873	0.5991
0.898	15.0	7935	1.9688	0.5976
0.8559	16.0	8464	1.9724	0.5988
0.8348	17.0	8993	1.9815	0.5714
0.8106	18.0	9522	2.0386	0.598
0.7848	19.0	10051	2.0627	0.5964
0.745	20.0	10580	2.0825	0.5966

Framework versions

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