Sneka's picture
End of training
cacefeb
metadata
license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
  - generated_from_trainer
model-index:
  - name: mistral-json-finetune
    results: []

mistral-json-finetune

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

  • Loss: 0.3730

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: 2.5e-05
  • train_batch_size: 2
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 5
  • training_steps: 1000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
1.2986 0.57 50 0.6897
0.4469 1.14 100 0.4239
0.2978 1.71 150 0.3497
0.2279 2.29 200 0.3126
0.1869 2.86 250 0.2858
0.137 3.43 300 0.2848
0.1221 4.0 350 0.2713
0.0918 4.57 400 0.2750
0.0832 5.14 450 0.2781
0.0681 5.71 500 0.2807
0.055 6.29 550 0.2902
0.0515 6.86 600 0.2746
0.0463 7.43 650 0.3078
0.0421 8.0 700 0.2971
0.0354 8.57 750 0.3196
0.0276 9.14 800 0.3490
0.0281 9.71 850 0.3411
0.026 10.29 900 0.3658
0.0241 10.86 950 0.3661
0.0209 11.43 1000 0.3730

Framework versions

  • Transformers 4.36.0.dev0
  • Pytorch 2.1.1+cu121
  • Datasets 2.15.0
  • Tokenizers 0.15.0