Edit model card

zephyr-Me

This model is a fine-tuned version of TheBloke/zephyr-7B-beta-GPTQ on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.0107
  • Rouge1: 0.7127
  • Rouge2: 0.4797
  • Rougel: 0.6694
  • Rougelsum: 0.6951
  • Meteor: 0.7003
  • F1 Score: 0.0010

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: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • num_epochs: 3
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Meteor F1 Score
2.1273 0.15 4 2.0726 0.4931 0.1866 0.4166 0.4753 0.4834 0.0077
1.6907 0.3 8 1.5193 0.6035 0.3389 0.5483 0.5905 0.5816 0.0010
1.3096 0.44 12 1.3236 0.6571 0.4159 0.6080 0.6386 0.6473 0.0008
1.1588 0.59 16 1.2651 0.6652 0.4210 0.6174 0.6455 0.6528 0.0008
1.1038 0.74 20 1.1852 0.6772 0.4239 0.6274 0.6557 0.6570 0.0008
1.0362 0.89 24 1.1448 0.6750 0.4256 0.6278 0.6547 0.6613 0.0008
1.0733 1.04 28 1.1137 0.6864 0.4397 0.6379 0.6655 0.6743 0.0008
0.8783 1.19 32 1.1179 0.6914 0.4510 0.6430 0.6680 0.6813 0.0010
0.8761 1.33 36 1.1020 0.6984 0.4545 0.6497 0.6768 0.6865 0.0010
0.8774 1.48 40 1.0696 0.7033 0.4604 0.6549 0.6834 0.6908 0.0010
0.8621 1.63 44 1.0485 0.7030 0.4642 0.6568 0.6850 0.6915 0.0010
0.8143 1.78 48 1.0334 0.7064 0.4670 0.6601 0.6874 0.6929 0.0010
0.7483 1.93 52 1.0232 0.7060 0.4681 0.6606 0.6868 0.6940 0.0010
0.7647 2.07 56 1.0148 0.7058 0.4700 0.6623 0.6884 0.6886 0.0010
0.6659 2.22 60 1.0135 0.7088 0.4737 0.6655 0.6917 0.6952 0.0010
0.7135 2.37 64 1.0098 0.7132 0.4783 0.6699 0.6948 0.6989 0.0010
0.6685 2.52 68 1.0123 0.7116 0.4787 0.6687 0.6939 0.6995 0.0010
0.6538 2.67 72 1.0113 0.7145 0.4811 0.6705 0.6966 0.7030 0.0010
0.6648 2.81 76 1.0108 0.7132 0.4800 0.6694 0.6955 0.7011 0.0010
0.6278 2.96 80 1.0107 0.7127 0.4797 0.6694 0.6951 0.7003 0.0010

Framework versions

  • PEFT 0.7.1
  • Transformers 4.35.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.0
  • Tokenizers 0.15.0
Downloads last month
75
Inference API
Unable to determine this model’s pipeline type. Check the docs .

Model tree for HimashaJ96/zephyr-Me

Adapter
(39)
this model