qlora-v1 / README.md
Laurie's picture
Create README.md
8049af9
|
raw
history blame
336 Bytes

model_name_or_path = yahma/llama-13b-hf learning_rate = 1e-4 max_steps = 10000 epoch = 3.08 train_loss = 0.7169 train_runtime = 1 day, 7:23:32.67 train_samples_per_second = 1.416 train_steps_per_second = 0.088