nelkh's picture
nelkh/pgd_mistral_8bits_lr0.0002_alpha32_rk4_do0.1_wd1.0e-02
17dc745 verified
metadata
base_model: mistralai/Mistral-7B-Instruct-v0.3
library_name: peft
license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: pgd_mistral_8bits_lr0.0002_alpha32_rk4_do0.1_wd1.0e-02
    results: []

pgd_mistral_8bits_lr0.0002_alpha32_rk4_do0.1_wd1.0e-02

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7793

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: 3
  • eval_batch_size: 3
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 12
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
3.0296 0.9778 11 2.3415
1.8108 1.9556 22 1.1645
1.0071 2.9333 33 0.8704
0.7805 4.0 45 0.8074
0.7974 4.9778 56 0.7788
0.7645 5.9556 67 0.7678
0.7508 6.9333 78 0.7640
0.675 8.0 90 0.7635
0.7262 8.9778 101 0.7708
0.6817 9.7778 110 0.7793

Framework versions

  • PEFT 0.10.1.dev0
  • Transformers 4.43.4
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1