Story Gen Results.txt

Basic:

TrainOutput(global_step=8145, training_loss=1.4378625278461192, metrics={'train_runtime': 7758.5612, 'train_samples_per_second': 67.195, 'train_steps_per_second': 1.05, 'total_flos': 7.054339513044173e+16, 'train_loss': 1.4378625278461192, 'epoch': 15.0})

{'eval_loss': 1.360207200050354,
 'eval_runtime': 19.5825,
 'eval_samples_per_second': 197.217,
 'eval_steps_per_second': 24.665,
 'epoch': 15.0}

3.98

LoRA:

TrainOutput(global_step=8145, training_loss=1.6368920744819242, metrics={'train_runtime': 7028.0123, 'train_samples_per_second': 74.18, 'train_steps_per_second': 1.159, 'total_flos': 7.101560826259046e+16, 'train_loss': 1.6368920744819242, 'epoch': 15.0})

{'eval_loss': 1.4845715761184692,
 'eval_runtime': 21.5247,
 'eval_samples_per_second': 179.421,
 'eval_steps_per_second': 22.439,
 'epoch': 15.0}

4.41

Adapters:

TrainOutput(global_step=8145, training_loss=1.7210480373281873, metrics={'train_runtime': 6142.2723, 'train_samples_per_second': 84.877, 'train_steps_per_second': 1.326, 'total_flos': 7.118346527440896e+16, 'train_loss': 1.7210480373281873, 'epoch': 15.0})

{'eval_loss': 1.5473278760910034,
 'eval_runtime': 19.6496,
 'eval_samples_per_second': 196.544,
 'eval_steps_per_second': 24.581,
 'epoch': 15.0}

4.69

DistilGPT2 - 15 epochs

global step = 8125
training loss = 2.1491
train_time = 56632
epoch = 15

{'eval_loss': 3.0107593536376953,
 'eval_runtime': 18.2229,
 'eval_samples_per_second': 211.931,
 'eval_steps_per_second': 26.505}

20.30

DistilGPT2 - 30 epochs

TrainOutput(global_step=16290, training_loss=1.0405259387049666, metrics={'train_runtime': 56691.3036, 'train_samples_per_second': 18.392, 'train_steps_per_second': 0.287, 'total_flos': 1.5087687350530867e+17, 'train_loss': 1.0405259387049666, 'epoch': 30.0})

{'eval_loss': 3.186328649520874,
 'eval_runtime': 21.7643,
 'eval_samples_per_second': 177.446,
 'eval_steps_per_second': 22.192}

24.20