Edit model card

data/openllama-3b-basque

This model is a fine-tuned version of openlm-research/open_llama_3b_v2 on the xezpeleta/oasst1_eu_threads dataset. It achieves the following results on the evaluation set:

  • Loss: 2.1151

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: 3e-06
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 20
  • num_epochs: 4

Training results

Training Loss Epoch Step Validation Loss
3.4011 0.0 1 3.5208
2.4173 0.25 917 2.6480
2.4849 0.5 1834 2.4465
2.2639 0.75 2751 2.3302
2.4403 1.0 3668 2.2548
2.1016 1.23 4585 2.2111
2.0178 1.48 5502 2.1805
2.041 1.73 6419 2.1476
2.0026 1.98 7336 2.1347
1.7586 2.22 8253 2.1187
1.8564 2.47 9170 2.1191
1.766 2.72 10087 2.1147
2.0431 2.97 11004 2.1157
2.3418 3.2 11921 2.1147
1.9713 3.45 12838 2.1151
1.9072 3.7 13755 2.1151

Framework versions

  • Transformers 4.37.0.dev0
  • Pytorch 2.0.1+cu118
  • Datasets 2.16.1
  • Tokenizers 0.15.0
Downloads last month
10

Finetuned from

Dataset used to train xezpeleta/openllama-3b-basque