Edit model card

phi-1_5-psychology

This model is a fine-tuned version of microsoft/phi-1_5 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7574

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: cosine
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss
0.8667 0.04 100 0.8554
0.8401 0.09 200 0.8524
0.8492 0.13 300 0.8437
0.8563 0.18 400 0.8393
0.8353 0.22 500 0.8367
0.8232 0.26 600 0.8305
0.8299 0.31 700 0.8226
0.8307 0.35 800 0.8233
0.8087 0.39 900 0.8170
0.8124 0.44 1000 0.8160
0.7943 0.48 1100 0.8103
0.7924 0.53 1200 0.8076
0.7918 0.57 1300 0.8026
0.807 0.61 1400 0.8012
0.788 0.66 1500 0.8034
0.7946 0.7 1600 0.7946
0.7959 0.75 1700 0.7926
0.7878 0.79 1800 0.7921
0.754 0.83 1900 0.7890
0.7762 0.88 2000 0.7850
0.7651 0.92 2100 0.7849
0.7868 0.97 2200 0.7855
0.7651 1.01 2300 0.7820
0.7323 1.05 2400 0.7818
0.7316 1.1 2500 0.7804
0.7311 1.14 2600 0.7808
0.7221 1.18 2700 0.7782
0.722 1.23 2800 0.7736
0.7217 1.27 2900 0.7780
0.7226 1.32 3000 0.7730
0.7305 1.36 3100 0.7731
0.7237 1.4 3200 0.7712
0.7127 1.45 3300 0.7710
0.7252 1.49 3400 0.7699
0.7076 1.54 3500 0.7687
0.7185 1.58 3600 0.7672
0.6921 1.62 3700 0.7639
0.6882 1.67 3800 0.7642
0.7184 1.71 3900 0.7633
0.7048 1.76 4000 0.7601
0.7136 1.8 4100 0.7598
0.7063 1.84 4200 0.7591
0.7054 1.89 4300 0.7589
0.6945 1.93 4400 0.7564
0.6955 1.97 4500 0.7544
0.6869 2.02 4600 0.7536
0.6477 2.06 4700 0.7566
0.6593 2.11 4800 0.7568
0.6441 2.15 4900 0.7562
0.6527 2.19 5000 0.7574

Framework versions

  • Transformers 4.33.2
  • Pytorch 2.0.1+cu118
  • Datasets 2.14.5
  • Tokenizers 0.13.3
Downloads last month
1
Inference Examples
Inference API (serverless) does not yet support model repos that contain custom code.

Finetuned from