llama-3-qlora-wizard-processed-indicator-0.6

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on the yihanwang617/WizardLM_70k_processed_indicator_unfiltered_4k dataset. It achieves the following results on the evaluation set:

• Loss: 0.6652

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.718	0.2225	200	0.7090
0.7205	0.4450	400	0.6897
0.7203	0.6675	600	0.6808
0.703	0.8900	800	0.6756
0.6759	1.1125	1000	0.6748
0.6533	1.3350	1200	0.6695
0.6458	1.5575	1400	0.6669
0.632	1.7800	1600	0.6655

Framework versions

• PEFT 0.12.0
• Transformers 4.40.1
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1