Leon97ZJU/llama_answer

llama

This model is a fine-tuned version of /inspire/hdd/ws-8207e9e2-e733-4eec-a475-cfa1c36480ba/embodied-multimodality/public/yli/workspace/Model/meta-llama/Llama-3.1-8B on the 2wikimultihopqa_train, the hotpotqa_train and the musique_train datasets. It achieves the following results on the evaluation set:

• Loss: 0.1784

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 8
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2074	0.1921	500	0.1999
0.2202	0.3843	1000	0.1942
0.1933	0.5764	1500	0.1811
0.1909	0.7686	2000	0.1720
0.1637	0.9607	2500	0.1639
0.0983	1.1528	3000	0.1664
0.1033	1.3450	3500	0.1602
0.1145	1.5371	4000	0.1557
0.105	1.7293	4500	0.1506
0.1008	1.9214	5000	0.1454
0.031	2.1136	5500	0.1753
0.0313	2.3057	6000	0.1777
0.0323	2.4978	6500	0.1765
0.0311	2.6900	7000	0.1781
0.0277	2.8821	7500	0.1787

Framework versions

• Transformers 4.46.1
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3