shawgpt-ft / README.md
minhdang's picture
minhdang/test-qlora
9ef61b9 verified
metadata
library_name: peft
tags:
  - generated_from_trainer
base_model: hanspeterlyngsoeraaschoujensen/deepseek-math-7b-instruct-GPTQ
model-index:
  - name: shawgpt-ft
    results: []

shawgpt-ft

This model is a fine-tuned version of hanspeterlyngsoeraaschoujensen/deepseek-math-7b-instruct-GPTQ on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.3960

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: 4
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 2
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
4.0223 0.9231 3 3.4527
3.7325 1.8462 6 3.1298
3.3119 2.7692 9 2.7966
2.1502 4.0 13 2.3471
2.4352 4.9231 16 2.0116
2.0017 5.8462 19 1.7512
1.6954 6.7692 22 1.5629
1.162 8.0 26 1.4357
1.4624 8.9231 29 1.3998
1.0468 9.2308 30 1.3960

Framework versions

  • PEFT 0.11.1
  • Transformers 4.40.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1