Buckets:
| import unittest | |
| import torch | |
| from training.detection_trainer import freeze_batchnorm_statistics, batchnorm_buffers | |
| class BatchNormPolicyTests(unittest.TestCase): | |
| def test_statistics_fixed_while_affine_parameters_learn(self): | |
| model = torch.nn.Sequential(torch.nn.BatchNorm2d(3), torch.nn.Conv2d(3, 1, 1)) | |
| before = batchnorm_buffers(model) | |
| affine = model[0].weight.detach().clone() | |
| optimizer = torch.optim.SGD(model.parameters(), lr=.1) | |
| for _ in range(2): | |
| model.train() | |
| freeze_batchnorm_statistics(model) | |
| optimizer.zero_grad() | |
| model(torch.ones(1, 3, 4, 4)).square().mean().backward() | |
| self.assertIsNotNone(model[0].weight.grad) | |
| optimizer.step() | |
| self.assertTrue(all(torch.equal(v, batchnorm_buffers(model)[k]) for k, v in before.items())) | |
| self.assertFalse(torch.equal(affine, model[0].weight)) | |
Xet Storage Details
- Size:
- 924 Bytes
- Xet hash:
- f83ec0b2a90e29f3a0e7f20fca69f82dd495170d8a2b087f297c3b50ff76303f
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.