llama-3-8B-lora-sft

This model is a fine-tuned version of ytcheng/llama-3-8B-pretrain on the sm_question dataset. It achieves the following results on the evaluation set:

• Loss: 1.9779

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.5512	0.1461	200	2.5884
2.3861	0.2923	400	2.4773
2.5938	0.4384	600	2.3990
2.3384	0.5845	800	2.3486
2.3232	0.7307	1000	2.3094
2.2603	0.8768	1200	2.2487
2.1007	1.0229	1400	2.2064
2.0635	1.1691	1600	2.1665
2.1212	1.3152	1800	2.1353
2.038	1.4613	2000	2.0880
1.9656	1.6075	2200	2.0618
2.0285	1.7536	2400	2.0393
1.9932	1.8997	2600	2.0270
1.6477	2.0458	2800	2.0164
1.5985	2.1920	3000	2.0024
1.8708	2.3381	3200	1.9918
1.8533	2.4842	3400	1.9873
1.7357	2.6304	3600	1.9813
1.8041	2.7765	3800	1.9782
1.7998	2.9226	4000	1.9780

Framework versions

• PEFT 0.10.0
• Transformers 4.40.0
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1