rheubanks's picture
rheubanks/LLMENG2
8cd2c12 verified
metadata
license: apache-2.0
library_name: peft
tags:
  - trl
  - sft
  - generated_from_trainer
datasets:
  - generator
base_model: mistralai/Mistral-7B-v0.1
model-index:
  - name: mistral7b_instruct_generation
    results: []

mistral7b_instruct_generation

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the generator dataset. It achieves the following results on the evaluation set:

  • Loss: 1.7939

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: 4
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: constant
  • lr_scheduler_warmup_steps: 0.03
  • training_steps: 500

Training results

Training Loss Epoch Step Validation Loss
1.7815 0.0 20 1.8296
1.8199 0.01 40 1.7992
1.7606 0.01 60 1.7950
1.8347 0.01 80 1.7923
1.7433 0.01 100 1.7958
1.8829 0.02 120 1.7928
1.769 0.02 140 1.7893
1.7817 0.02 160 1.7849
1.7975 0.03 180 1.7881
2.008 0.03 200 1.7882
1.827 0.03 220 1.7993
1.8336 0.03 240 1.7953
1.8757 0.04 260 1.7916
1.9317 0.04 280 1.7900
1.8708 0.04 300 1.7867
1.8851 0.04 320 1.7928
1.94 0.05 340 1.7880
1.7749 0.05 360 1.8033
1.8647 0.05 380 1.7870
1.8468 0.06 400 1.7871
1.8341 0.06 420 1.7890
1.9152 0.06 440 1.7892
1.7979 0.06 460 1.8051
1.9065 0.07 480 1.7986
1.8011 0.07 500 1.7939

Framework versions

  • PEFT 0.7.1
  • Transformers 4.37.0
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0