qwerrwe / tests /test_schedulers.py
jinwonkim93's picture
Scheduler implementation of Continual Pre-Training of Large Language Models: How to (re)warm your model? (#1273)
8430db2 unverified
raw
history blame
1.72 kB
"""
test module for the axolotl.utis.data module
"""
import unittest
import torch
from torch.optim import SGD
from axolotl.utils.schedulers import get_cosine_schedule_with_warmup_decay_constant
class TestCosineConstantLr(unittest.TestCase):
"""
test class for encode pretraining and md5 helper
"""
def setUp(self):
self.train_steps = 1000
self.warmup_steps = 10
self.min_lr_ratio = 0.1
self.constant_lr_ratio = 0.8
self._lr = 0.01
self.optimizer = SGD([torch.tensor(1)], lr=self._lr)
self.lr_scheduler = get_cosine_schedule_with_warmup_decay_constant( # pylint: disable=attribute-defined-outside-init
self.optimizer,
num_warmup_steps=self.warmup_steps,
num_training_steps=self.train_steps,
min_lr_ratio=self.min_lr_ratio,
constant_lr_ratio=self.constant_lr_ratio,
)
def test_schedulers(self):
self.assertEqual(self.lr_scheduler.get_last_lr()[0], 0)
for _ in range(self.warmup_steps):
self.lr_scheduler.step()
self.assertEqual(self.lr_scheduler.get_last_lr()[0], self._lr)
constant_step = int(self.train_steps * self.constant_lr_ratio)
remaining_step = self.train_steps - constant_step
for _ in range(constant_step):
self.lr_scheduler.step()
self.assertEqual(
self.lr_scheduler.get_last_lr()[0], self._lr * self.min_lr_ratio
)
for _ in range(remaining_step):
self.lr_scheduler.step()
self.assertEqual(
self.lr_scheduler.get_last_lr()[0], self._lr * self.min_lr_ratio
)
if __name__ == "__main__":
unittest.main()