Merge branch 'dev' into origin/main

.github/workflows/python-test.yml

@@ -16,7 +16,7 @@ jobs:
     strategy:
       matrix:
         os: [ubuntu-latest] # [ubuntu-latest, windows-latest, macos-latest]
-        python-version: ["3.8", "3.9"]
+        python-version: ["3.8", "3.9", "3.10"]
 
     steps:
     - uses: actions/checkout@v2

disent/frameworks/vae/_weaklysupervised__adavae.py

@@ -218,7 +218,7 @@ def compute_average_gvae_std(d0_posterior: Normal, d1_posterior: Normal) -> Norm
     assert isinstance(d1_posterior, Normal), f'posterior distributions must be {Normal.__name__} distributions, got: {type(d1_posterior)}'
     # averages
     ave_std = 0.5 * (d0_posterior.stddev + d1_posterior.stddev)
-    ave_mean = 0.5 * (d1_posterior.mean + d1_posterior.mean)
+    ave_mean = 0.5 * (d0_posterior.mean + d1_posterior.mean)
     # done!
     return Normal(loc=ave_mean, scale=ave_std)
 
@@ -235,7 +235,7 @@ def compute_average_gvae(d0_posterior: Normal, d1_posterior: Normal) -> Normal:
     assert isinstance(d1_posterior, Normal), f'posterior distributions must be {Normal.__name__} distributions, got: {type(d1_posterior)}'
     # averages
     ave_var = 0.5 * (d0_posterior.variance + d1_posterior.variance)
-    ave_mean = 0.5 * (d1_posterior.mean + d1_posterior.mean)
+    ave_mean = 0.5 * (d0_posterior.mean + d1_posterior.mean)
     # done!
     return Normal(loc=ave_mean, scale=torch.sqrt(ave_var))
 
@@ -323,10 +323,10 @@ def hook_intercept_ds(self, ds_posterior: Sequence[Distribution], ds_prior: Sequ
         ave_std  = (0.5 * d0_posterior.variance + 0.5 * d1_posterior.variance) ** 0.5
 
         # [4.b] select shared or original values based on mask
-        z0_mean = torch.where(share_mask,  d0_posterior.loc,   ave_mean)
-        z1_mean = torch.where(share_mask,  d1_posterior.loc,   ave_mean)
-        z0_std  = torch.where(share_mask,  d0_posterior.scale, ave_std)
-        z1_std  = torch.where(share_mask,  d1_posterior.scale, ave_std)
+        z0_mean = torch.where(share_mask, ave_mean, d0_posterior.loc)
+        z1_mean = torch.where(share_mask, ave_mean, d1_posterior.loc)
+        z0_std  = torch.where(share_mask, ave_std,  d0_posterior.scale)
+        z1_std  = torch.where(share_mask, ave_std,  d1_posterior.scale)
 
         # construct distributions
         ave_d0_posterior = Normal(loc=z0_mean, scale=z0_std)

disent/util/seeds.py

@@ -24,8 +24,6 @@
 
 import contextlib
 import logging
-import random
-import numpy as np
 
 
 log = logging.getLogger(__name__)
@@ -44,8 +42,10 @@ def seed(long=777):
         log.warning(f'[SEEDING]: no seed was specified. Seeding skipped!')
         return
     # seed python
+    import random
     random.seed(long)
     # seed numpy
+    import numpy as np
     np.random.seed(long)
     # seed torch - it can be slow to import
     try:
@@ -60,34 +60,34 @@ def seed(long=777):
 
 
 class TempNumpySeed(contextlib.ContextDecorator):
-    def __init__(self, seed=None, offset=0):
+    def __init__(self, seed: int = None):
         # check and normalize seed
         if seed is not None:
             try:
                 seed = int(seed)
             except:
-                raise ValueError(f'{seed=} is not int-like!')
-        # offset seed
-        if seed is not None:
-            seed += offset
+                raise ValueError(f'seed={seed} is not int-like!')
         # save values
         self._seed = seed
         self._state = None
 
     def __enter__(self):
         if self._seed is not None:
+            import numpy as np
             self._state = np.random.get_state()
             np.random.seed(self._seed)
 
     def __exit__(self, *args, **kwargs):
         if self._seed is not None:
+            import numpy as np
             np.random.set_state(self._state)
             self._state = None
 
     def _recreate_cm(self):
         # TODO: do we need to override this?
         return self
 
+
 # ========================================================================= #
 # END                                                                       #
 # ========================================================================= #

setup.py

@@ -48,7 +48,7 @@
     author="Nathan Juraj Michlo",
     author_email="[email protected]",
 
-    version="0.5.0",
+    version="0.5.1",
     python_requires=">=3.8",  # we make use of standard library features only in 3.8
     packages=setuptools.find_packages(),
 
@@ -64,6 +64,7 @@
         "Operating System :: OS Independent",
         "Programming Language :: Python :: 3.8",
         "Programming Language :: Python :: 3.9",
+        "Programming Language :: Python :: 3.10",
         "Intended Audience :: Science/Research",
     ],
 )

tests/test_frameworks.py

@@ -28,13 +28,15 @@
 
 import pytest
 import pytorch_lightning as pl
+import torch
 from torch.utils.data import DataLoader
 
 from disent.dataset import DisentDataset
 from disent.dataset.data import XYObjectData
 from disent.dataset.sampling import GroundTruthSingleSampler
 from disent.dataset.sampling import GroundTruthPairSampler
 from disent.dataset.sampling import GroundTruthTripleSampler
+from disent.dataset.sampling import RandomSampler
 from disent.frameworks.ae import *
 from disent.frameworks.vae import *
 from disent.model import AutoEncoder
@@ -46,6 +48,8 @@
 # ========================================================================= #
 # TEST FRAMEWORKS                                                           #
 # ========================================================================= #
+from disent.util.seeds import seed
+from disent.util.seeds import TempNumpySeed
 from docs.examples.extend_experiment.code.weaklysupervised__si_adavae import SwappedInputAdaVae
 from docs.examples.extend_experiment.code.weaklysupervised__si_betavae import SwappedInputBetaVae
 
@@ -166,6 +170,53 @@ def test_framework_config_defaults():
     )
 
 
+def test_ada_vae_similarity():
+
+    seed(42)
+
+    data = XYObjectData()
+    dataset = DisentDataset(data, sampler=RandomSampler(num_samples=2), transform=ToImgTensorF32())
+    dataloader = DataLoader(dataset, num_workers=0, batch_size=3)
+
+    model = AutoEncoder(
+        encoder=EncoderLinear(x_shape=data.x_shape, z_size=25, z_multiplier=2),
+        decoder=DecoderLinear(x_shape=data.x_shape, z_size=25, z_multiplier=1),
+    )
+
+    adavae0 = AdaGVaeMinimal(model=model, cfg=AdaGVaeMinimal.cfg())
+    adavae1 = AdaVae(model=model, cfg=AdaVae.cfg())
+    adavae2 = AdaVae(model=model, cfg=AdaVae.cfg(
+        ada_average_mode='gvae',
+        ada_thresh_mode='symmetric_kl',
+        ada_thresh_ratio=0.5,
+    ))
+
+    batch = next(iter(dataloader))
+
+    # TODO: add a TempNumpySeed equivalent for torch
+    seed(777)
+    result0a = adavae0.do_training_step(batch, 0)
+    seed(777)
+    result0b = adavae0.do_training_step(batch, 0)
+    assert torch.allclose(result0a, result0b), f'{result0a} does not match {result0b}'
+
+    seed(777)
+    result1a = adavae1.do_training_step(batch, 0)
+    seed(777)
+    result1b = adavae1.do_training_step(batch, 0)
+    assert torch.allclose(result1a, result1b), f'{result1a} does not match {result1b}'
+
+    seed(777)
+    result2a = adavae2.do_training_step(batch, 0)
+    seed(777)
+    result2b = adavae2.do_training_step(batch, 0)
+    assert torch.allclose(result2a, result2b), f'{result2a} does not match {result2b}'
+
+    # check similar
+    assert torch.allclose(result0a, result1a), f'{result0a} does not match {result1a}'
+    assert torch.allclose(result1a, result2a), f'{result1a} does not match {result2a}'
+
+
 # ========================================================================= #
 # END                                                                       #
 # ========================================================================= #