File size: 3,852 Bytes
94f7dc4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 | import collections
import unittest
from multisource_core import license_header,code_records,evaluation_record,units,doc_records
META={'source_id':'test_source','source_repo':'org/repo','source_revision':'abc',
'source_path':'model.py','source_url':'https://example.org/model.py','source_group':'example','source_row':-1,'upstream_track':'main'}
class MultiSourceTests(unittest.TestCase):
def test_unknown_license_is_not_inferred_from_mit_dataset_wrapper(self):
counters=collections.Counter()
self.assertFalse(code_records('import torch\nx=torch.tensor([1,2])',META,counters,default_license='MIT',require_inline=True))
self.assertEqual(counters['missing_original_license_evidence'],1)
def test_explicit_spdx_is_preserved(self):
code='# Copyright (c) Example Author\n# SPDX-License-Identifier: Apache-2.0\nimport torch\nx=torch.tensor([1,2])'
license_id,notice,evidence=license_header(code)
self.assertEqual(license_id,'Apache-2.0');self.assertIn('Example Author',notice)
self.assertEqual(evidence,'inline_spdx')
def test_noncommercial_does_not_become_mit(self):
code='# Copyright NVIDIA\n# Nvidia Source Code License-NC\nimport torch\nx=torch.tensor([1,2])'
self.assertEqual(license_header(code)[2],'restricted_or_unsupported_header')
def test_root_bsd_repository_license_is_used_only_without_conflict(self):
rows=code_records('import torch\nx=torch.tensor([1,2])',META,collections.Counter(),default_license='BSD-3-Clause')
self.assertTrue(rows);self.assertEqual(rows[0]['license_id'],'BSD-3-Clause')
def test_cnn_survives_unsafe_download_entrypoint(self):
code='''import torch
from torch import nn
from torchvision import datasets
class Net(nn.Module):
def __init__(self):
super().__init__()
self.conv=nn.Conv2d(3,8,3)
def forward(self,x):
return self.conv(x)
if __name__=='__main__':
datasets.CIFAR10('.',download=True)
'''
rows=code_records(code,META,collections.Counter(),default_license='BSD-3-Clause')
self.assertTrue(any('cnn_from_scratch_candidate' in r['topics'] for r in rows))
self.assertTrue(all('download=True' not in r['code'] for r in rows))
def test_decorator_is_not_lost(self):
code='import torch\n@torch.no_grad()\ndef predict(x):\n return torch.softmax(x,dim=-1)'
rows=code_records(code,META,collections.Counter(),default_license='BSD-3-Clause')
self.assertTrue(rows);self.assertIn('@torch.no_grad()',rows[0]['code'])
def test_guarded_import_is_never_hoisted(self):
code='try:\n import optional_module\nexcept ImportError:\n pass\nimport torch\ndef f(x):\n return torch.relu(x)'
rows=code_records(code,META,collections.Counter(),default_license='BSD-3-Clause')
self.assertTrue(rows);self.assertNotIn('optional_module',rows[-1]['code'])
def test_benchmark_has_no_training_split_or_recipe_claim(self):
r=evaluation_record(META,'case1','An intentionally failing shape problem.','MIT','import torch\nx=torch.tensor([1,2])')
self.assertTrue(r['benchmark_content']);self.assertEqual(r['split'],'test');self.assertEqual(r['intended_use'],'evaluation')
self.assertIn('not a working recipe',r['text'])
def test_secret_and_process_calls_rejected(self):
rows=code_records("import subprocess\nsubprocess.run(['whoami'])",META,collections.Counter(),default_license='MIT')
self.assertFalse(rows)
def test_documentation_does_not_split_python_fences(self):
text='PyTorch tensor training documentation. '*12+'\n\n```python\nprint(123)\n```\n\nMore PyTorch explanation. '*8
rows=doc_records(text,META,'BSD-3-Clause',collections.Counter())
self.assertTrue(rows);self.assertTrue(all('```' not in r['text'] for r in rows))
if __name__=='__main__':unittest.main()
|