llama-3-86-lora-pretrain_v2

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the sm_artile dataset. It achieves the following results on the evaluation set:

• Loss: 2.2353

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
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• total_eval_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.8142	0.1877	100	2.8566
2.7235	0.3755	200	2.6776
2.5809	0.5632	300	2.5664
2.3971	0.7510	400	2.4458
2.4147	0.9387	500	2.3812
2.3987	1.1265	600	2.3436
2.3	1.3142	700	2.3193
2.3219	1.5020	800	2.2951
2.377	1.6897	900	2.2763
2.2977	1.8775	1000	2.2623
2.269	2.0652	1100	2.2525
2.2305	2.2530	1200	2.2442
2.3866	2.4407	1300	2.2396
2.3217	2.6285	1400	2.2369
2.2007	2.8162	1500	2.2355

Framework versions

• PEFT 0.10.0
• Transformers 4.40.0
• Pytorch 2.2.1
• Datasets 2.18.0
• Tokenizers 0.19.1