Hugging Face's logo

system-k
/
gemma-2-9b-evaluator

PEFT
Safetensors
Generated from Trainer
Model card Files Community
gemma-2-9b-evaluator / README.md
naive-puzzle's picture
naive-puzzle
End of training
e077085 verified
preview code
|
raw
history blame contribute delete
14.6 kB
metadata
library_name: peft
license: gemma
base_model: google/gemma-2-9b-it
tags:
  - generated_from_trainer
model-index:
  - name: gemma-2-9b-evaluator-v1
    results: []

gemma-2-9b-evaluator-v1

This model is a fine-tuned version of google/gemma-2-9b-it on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.6849
  • Helpfulness Accuracy: 0.4181
  • Helpfulness Spearmanr: 0.4129
  • Helpfulness Kendalltau: 0.3194
  • Helpfulness Pearsonr: 0.5143
  • Helpfulness Rmse: 1.0735
  • Helpfulness Mae: 0.8559
  • Correctness Accuracy: 0.4769
  • Correctness Spearmanr: 0.4135
  • Correctness Kendalltau: 0.3214
  • Correctness Pearsonr: 0.5143
  • Correctness Rmse: 1.0616
  • Correctness Mae: 0.8366
  • Coherence Accuracy: 0.7148
  • Coherence Spearmanr: 0.2880
  • Coherence Kendalltau: 0.2322
  • Coherence Pearsonr: 0.3705
  • Coherence Rmse: 0.6259
  • Coherence Mae: 0.4798
  • Complexity Accuracy: 0.6012
  • Complexity Spearmanr: 0.5174
  • Complexity Kendalltau: 0.4180
  • Complexity Pearsonr: 0.5310
  • Complexity Rmse: 0.6137
  • Complexity Mae: 0.4678
  • Verbosity Accuracy: 0.6763
  • Verbosity Spearmanr: 0.6069
  • Verbosity Kendalltau: 0.4930
  • Verbosity Pearsonr: 0.6736
  • Verbosity Rmse: 0.5983
  • Verbosity Mae: 0.4485
  • Avg Accuracy: 0.5775
  • Avg Spearmanr: 0.4477
  • Avg Kendalltau: 0.3568
  • Avg Pearsonr: 0.5207
  • Avg Rmse: 0.7946
  • Avg Mae: 0.6177

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: 1e-05
  • train_batch_size: 2
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine_with_min_lr
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 1.0

Training results

Training Loss Epoch Step Validation Loss Helpfulness Accuracy Helpfulness Spearmanr Helpfulness Kendalltau Helpfulness Pearsonr Helpfulness Rmse Helpfulness Mae Correctness Accuracy Correctness Spearmanr Correctness Kendalltau Correctness Pearsonr Correctness Rmse Correctness Mae Coherence Accuracy Coherence Spearmanr Coherence Kendalltau Coherence Pearsonr Coherence Rmse Coherence Mae Complexity Accuracy Complexity Spearmanr Complexity Kendalltau Complexity Pearsonr Complexity Rmse Complexity Mae Verbosity Accuracy Verbosity Spearmanr Verbosity Kendalltau Verbosity Pearsonr Verbosity Rmse Verbosity Mae Avg Accuracy Avg Spearmanr Avg Kendalltau Avg Pearsonr Avg Rmse Avg Mae
No log 0 0 7.3447 0.0934 -0.1708 -0.1290 -0.1865 2.6016 2.3248 0.2630 -0.0067 -0.0052 -0.0083 1.3643 1.1593 0.0530 -0.0417 -0.0336 -0.0564 3.0362 2.8994 0.0983 -0.1112 -0.0871 -0.1231 1.0970 0.8836 0.1378 -0.0086 -0.0068 -0.0139 1.2755 1.0756 0.1291 -0.0678 -0.0523 -0.0776 1.8749 1.6686
2.1437 0.0719 500 2.3464 0.3092 0.0421 0.0320 0.0373 1.2641 1.0162 0.3198 0.1021 0.0767 0.1458 1.2286 1.0016 0.5491 0.0236 0.0188 0.0441 0.7506 0.6140 0.4711 0.1512 0.1182 0.1430 0.7263 0.5881 0.5588 0.1844 0.1419 0.2079 0.7855 0.5510 0.4416 0.1007 0.0775 0.1156 0.9510 0.7542
1.843 0.1437 1000 1.9347 0.3998 0.3252 0.2481 0.3956 1.1578 0.9084 0.4403 0.3335 0.2557 0.4149 1.1291 0.8900 0.6628 0.1423 0.1139 0.1817 0.6701 0.5221 0.5607 0.3808 0.3022 0.4127 0.6576 0.5173 0.6243 0.4424 0.3485 0.4902 0.7009 0.4784 0.5376 0.3248 0.2537 0.3790 0.8631 0.6632
1.8118 0.2156 1500 1.8294 0.3738 0.3565 0.2739 0.4501 1.1206 0.8954 0.4316 0.3606 0.2784 0.4580 1.0958 0.8865 0.6830 0.1907 0.1529 0.2532 0.6581 0.5247 0.5848 0.4564 0.3656 0.4838 0.6362 0.4909 0.6368 0.5178 0.4127 0.5913 0.6608 0.4474 0.5420 0.3764 0.2967 0.4473 0.8343 0.6490
1.6318 0.2875 2000 1.7885 0.3882 0.3734 0.2886 0.4616 1.1076 0.8827 0.4701 0.3800 0.2947 0.4699 1.0894 0.8631 0.6965 0.2344 0.1883 0.2992 0.6412 0.4804 0.5867 0.4633 0.3721 0.4818 0.6373 0.4734 0.6320 0.5412 0.4338 0.6172 0.6365 0.4765 0.5547 0.3985 0.3155 0.4659 0.8224 0.6352
1.7944 0.3594 2500 1.7645 0.4046 0.3903 0.3016 0.4828 1.0943 0.8715 0.4576 0.3930 0.3053 0.4839 1.0783 0.8590 0.6840 0.2530 0.2035 0.3156 0.6434 0.5073 0.5838 0.4831 0.3894 0.4970 0.6277 0.4767 0.6474 0.5542 0.4453 0.6269 0.6427 0.4815 0.5555 0.4147 0.3290 0.4812 0.8173 0.6392
1.6247 0.4312 3000 1.7417 0.3545 0.3994 0.3085 0.4981 1.0894 0.8818 0.4672 0.4033 0.3128 0.5024 1.0650 0.8560 0.7110 0.2636 0.2123 0.3346 0.6330 0.4812 0.5915 0.4970 0.4012 0.5094 0.6237 0.4733 0.6532 0.5731 0.4623 0.6450 0.6289 0.4780 0.5555 0.4273 0.3394 0.4979 0.8080 0.6341
1.5258 0.5031 3500 1.7147 0.3950 0.4007 0.3098 0.4954 1.0861 0.8657 0.4730 0.4060 0.3154 0.4983 1.0706 0.8456 0.7148 0.2700 0.2174 0.3492 0.6280 0.4644 0.6040 0.5005 0.4037 0.5149 0.6216 0.4902 0.6590 0.5813 0.4698 0.6510 0.6043 0.4418 0.5692 0.4317 0.3432 0.5018 0.8021 0.6215
1.6146 0.5750 4000 1.7121 0.4114 0.4005 0.3095 0.5034 1.0830 0.8597 0.4769 0.4057 0.3152 0.5055 1.0649 0.8371 0.7225 0.2730 0.2200 0.3536 0.6249 0.4447 0.5954 0.5037 0.4064 0.5112 0.6232 0.4644 0.6638 0.5981 0.4850 0.6605 0.6168 0.4499 0.5740 0.4362 0.3472 0.5068 0.8026 0.6112
1.517 0.6468 4500 1.7076 0.4171 0.4057 0.3134 0.5066 1.0792 0.8598 0.4807 0.4093 0.3179 0.5092 1.0628 0.8370 0.7216 0.2781 0.2240 0.3628 0.6219 0.4552 0.5915 0.5043 0.4070 0.5095 0.6316 0.4612 0.6667 0.6012 0.4878 0.6676 0.6101 0.4619 0.5755 0.4397 0.3500 0.5111 0.8011 0.6150
1.6201 0.7187 5000 1.6972 0.4143 0.4073 0.3146 0.5110 1.0756 0.8582 0.4836 0.4090 0.3176 0.5121 1.0610 0.8329 0.7197 0.2816 0.2271 0.3668 0.6214 0.4611 0.5944 0.5126 0.4142 0.5222 0.6223 0.4624 0.6647 0.6072 0.4931 0.6724 0.6066 0.4627 0.5753 0.4435 0.3533 0.5169 0.7974 0.6155
1.6377 0.7906 5500 1.6846 0.4220 0.4099 0.3168 0.5128 1.0754 0.8550 0.4798 0.4111 0.3196 0.5127 1.0591 0.8381 0.7245 0.2843 0.2292 0.3684 0.6206 0.4495 0.6021 0.5131 0.4145 0.5276 0.6144 0.4601 0.6696 0.6092 0.4951 0.6752 0.5966 0.4445 0.5796 0.4455 0.3550 0.5193 0.7932 0.6095
1.619 0.8624 6000 1.6815 0.4027 0.4127 0.3192 0.5166 1.0698 0.8549 0.4817 0.4141 0.3220 0.5166 1.0560 0.8366 0.7197 0.2834 0.2284 0.3670 0.6230 0.4638 0.6002 0.5170 0.4178 0.5325 0.6129 0.4619 0.6686 0.6071 0.4931 0.6729 0.5976 0.4475 0.5746 0.4468 0.3561 0.5211 0.7918 0.6129
1.7326 0.9343 6500 1.6849 0.4181 0.4129 0.3194 0.5143 1.0735 0.8559 0.4769 0.4135 0.3214 0.5143 1.0616 0.8366 0.7148 0.2880 0.2322 0.3705 0.6259 0.4798 0.6012 0.5174 0.4180 0.5310 0.6137 0.4678 0.6763 0.6069 0.4930 0.6736 0.5983 0.4485 0.5775 0.4477 0.3568 0.5207 0.7946 0.6177

Framework versions

  • PEFT 0.15.0
  • Transformers 4.50.0.dev0
  • Pytorch 2.4.1+cu124
  • Datasets 3.4.1
  • Tokenizers 0.21.1