lmind_hotpot_train8000_eval7405_v1_docidx_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_docidx dataset. It achieves the following results on the evaluation set:

• Loss: 0.7747
• Accuracy: 0.7938

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
1.1109	1.0	839	1.3437	0.7529
1.0723	2.0	1678	1.2988	0.7550
1.0453	3.0	2517	1.2332	0.7577
1.0099	4.0	3357	1.2146	0.7604
0.9553	5.0	4196	1.1671	0.7632
0.8876	6.0	5035	1.1263	0.7655
0.8352	7.0	5874	1.0776	0.7681
0.7872	8.0	6714	1.0745	0.7706
0.7297	9.0	7553	1.0479	0.7730
0.6831	10.0	8392	1.0078	0.7754
0.6397	11.0	9231	0.9763	0.7779
0.5885	12.0	10071	0.9702	0.7803
0.5379	13.0	10910	0.9445	0.7824
0.4996	14.0	11749	0.9087	0.7846
0.464	15.0	12588	0.8827	0.7866
0.4225	16.0	13428	0.8886	0.7881
0.4259	17.0	14267	0.8224	0.7898
0.361	18.0	15106	0.7985	0.7913
0.3429	19.0	15945	0.7804	0.7930
0.305	19.99	16780	0.7747	0.7938

Framework versions

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