chansung's picture
End of training
0eb3a97 verified
metadata
license: gemma
library_name: peft
tags:
  - alignment-handbook
  - trl
  - sft
  - generated_from_trainer
base_model: google/gemma-2b
datasets:
  - llama-duo/synth_summarize_dataset_dedup
model-index:
  - name: gemma2b-summarize-claude3sonnet-128k
    results: []

gemma2b-summarize-claude3sonnet-128k

This model is a fine-tuned version of google/gemma-2b on the llama-duo/synth_summarize_dataset_dedup dataset. It achieves the following results on the evaluation set:

  • Loss: 2.6928

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: 8
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 3
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 48
  • total_eval_batch_size: 24
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 10

Training results

Training Loss Epoch Step Validation Loss
1.0192 1.0 402 2.4514
0.9424 2.0 804 2.4604
0.8955 3.0 1206 2.5064
0.8659 4.0 1608 2.5306
0.8359 5.0 2010 2.5706
0.7986 6.0 2412 2.6196
0.7778 7.0 2814 2.6583
0.7562 8.0 3216 2.6846
0.7563 9.0 3618 2.6927
0.7461 10.0 4020 2.6928

Framework versions

  • PEFT 0.11.1
  • Transformers 4.41.2
  • Pytorch 2.2.2+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1