oodeel 0.1.1__py3-none-any.whl → 0.3.0__py3-none-any.whl

tests/tests_torch/tools_torch.py

@@ -35,14 +35,14 @@ def almost_equal(arr1, arr2, epsilon=1e-6):
 
 
 class Net(nn.Module):
-    def __init__(self):
+    def __init__(self, num_classes=10):
         super().__init__()
         self.conv1 = nn.Conv2d(3, 6, 5)
         self.pool = nn.MaxPool2d(2, 2)
         self.conv2 = nn.Conv2d(6, 16, 5)
         self.fc1 = nn.Linear(16 * 5 * 5, 120)
         self.fc2 = nn.Linear(120, 84)
-        self.fc3 = nn.Linear(84, 10)
+        self.fc3 = nn.Linear(84, num_classes)
 
     def forward(self, x):
         x = self.pool(F.relu(self.conv1(x)))
@@ -55,7 +55,7 @@ class Net(nn.Module):
 
 
 class ComplexNet(nn.Module):
-    def __init__(self):
+    def __init__(self, num_classes=10):
         super().__init__()
 
         self.feature_extractor = nn.Sequential(
@@ -77,7 +77,7 @@ class ComplexNet(nn.Module):
                 [
                     ("fc1", nn.Linear(16 * 5 * 5, 120)),
                     ("fc2", nn.Linear(120, 84)),
-                    ("fc3", nn.Linear(84, 10)),
+                    ("fc3", nn.Linear(84, num_classes)),
                 ]
             )
         )
@@ -88,7 +88,7 @@ class ComplexNet(nn.Module):
         return x
 
 
-def sequential_model():
+def sequential_model(num_classes=10):
     return nn.Sequential(
         nn.Conv2d(3, 6, 5),
         nn.ReLU(),
@@ -99,11 +99,11 @@ def sequential_model():
         nn.Flatten(),
         nn.Linear(16 * 5 * 5, 120),
         nn.Linear(120, 84),
-        nn.Linear(84, 10),
+        nn.Linear(84, num_classes),
     )
 
 
-def named_sequential_model():
+def named_sequential_model(num_classes=10):
     return nn.Sequential(
         OrderedDict(
             [
@@ -116,13 +116,13 @@ def named_sequential_model():
                 ("flatten", nn.Flatten()),
                 ("fc1", nn.Linear(16 * 5 * 5, 120)),
                 ("fc2", nn.Linear(120, 84)),
-                ("fc3", nn.Linear(84, 10)),
+                ("fc3", nn.Linear(84, num_classes)),
             ]
         )
     )
 
 
-def simplest_mlp(num_features, num_classes):
+def simplest_mlp(num_features, num_classes=10):
     return nn.Sequential(
         nn.Linear(num_features, 64),
         nn.ReLU(),

tests/tests_torch/torch_methods_utils.py

@@ -23,12 +23,11 @@
 import os
 
 import numpy as np
-import requests
 import torch
 from sklearn.datasets import make_blobs
 from sklearn.model_selection import train_test_split
 
-from oodeel.datasets import OODDataset
+from oodeel.datasets import load_data_handler
 from oodeel.eval.metrics import bench_metrics
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@@ -53,31 +52,26 @@ def load_blobs_data(batch_size=128, num_samples=10000, train_ratio=0.8):
     )
 
     # === id / ood split ===
-    blobs_train = OODDataset((X_train, y_train), backend="torch")
-    blobs_test = OODDataset((X_test, y_test), backend="torch")
-    oods_fit, _ = blobs_train.split_by_class(in_labels, out_labels)
-    oods_in, oods_out = blobs_test.split_by_class(in_labels, out_labels)
+    handler = load_data_handler("torch")
+    blobs_train = handler.load_dataset((X_train, y_train))
+    blobs_test = handler.load_dataset((X_test, y_test))
+    oods_fit, _ = handler.split_by_class(blobs_train, in_labels, out_labels)
+    oods_in, oods_out = handler.split_by_class(blobs_test, in_labels, out_labels)
 
     # === prepare data (shuffle, batch) => torch dataloaders ===
-    ds_fit = oods_fit.prepare(batch_size=batch_size, shuffle=True)
-    ds_in = oods_in.prepare(batch_size=batch_size)
-    ds_out = oods_out.prepare(batch_size=batch_size)
+    ds_fit = handler.prepare(oods_fit, batch_size=batch_size, shuffle=True)
+    ds_in = handler.prepare(oods_in, batch_size=batch_size)
+    ds_out = handler.prepare(oods_out, batch_size=batch_size)
     return ds_fit, ds_in, ds_out
 
 
 def load_blob_mlp():
-    model_path_blob = os.path.join(model_path, "blobs_mlp.pt")
-
-    # if model not in local, download it
-    if not os.path.exists(model_path_blob):
-        data = requests.get(
-            "https://share.deel.ai/s/xcyk3ET8fzfTp8S/download/blobs_mlp.pt"
-        )
-        with open(model_path_blob, "wb") as file:
-            file.write(data.content)
-
     # load model
-    model = torch.load(model_path_blob, map_location=device)
+    model = torch.hub.load_state_dict_from_url(
+        "https://github.com/deel-ai/oodeel/blob/assets/"
+        + "test_models/blobs_mlp.pt?raw=True",
+        map_location=device,
+    )
     model.eval()
     return model
 
@@ -125,14 +119,27 @@ def eval_detector_on_blobs(
 
     # react specific test
     # /!\ do it at the end of the test because it may affect the detector's behaviour
+
     if check_react_clipping:
         assert detector.react_threshold is not None
+
         penult_feat_extractor = detector._load_feature_extractor(
-            model=model, feature_layers_id=[-2, -1]
-        )
-        penult_features = penult_feat_extractor.predict(ds_fit)[0][0]
-        assert torch.max(penult_features) <= detector.react_threshold, (
-            f"Maximum value of penultimate features ({torch.max(penult_features)})"
-            + " should be less than or equal to the react threshold value"
-            + f" ({detector.react_threshold})"
+            model=model, feature_layers_id=[-1]
         )
+        # penult_feat_extractor._prepare_ood_handles()
+
+        def hook(_, input):
+            penult_feat_extractor._features["test"] = input[0]
+
+        penult_feat_extractor.model[-1].register_forward_pre_hook(hook)
+        for x, y in ds_fit:
+            _ = penult_feat_extractor.predict_tensor(x)
+            assert (
+                torch.max(penult_feat_extractor._features["test"])
+                <= detector.react_threshold
+            ), (
+                "Maximum value of penultimate features"
+                + f" ({torch.max(penult_feat_extractor._features)})"
+                + " should be less than or equal to the react threshold value"
+                + f" ({detector.react_threshold})"
+            )

tests/tools_operator.py

@@ -127,7 +127,7 @@ def check_common_operators(backend):
     assert operator.flatten(x).shape == (25, 12 * 6)
 
     # Transpose
-    assert operator.transpose(x[0]).shape == (6, 12)
+    assert operator.t(x[0]).shape == (6, 12)
 
     # Diag
     assert operator.diag(x[0]).shape == (6,)
@@ -142,3 +142,12 @@ def check_common_operators(backend):
 
     # Pinv
     assert operator.pinv(x[0]).shape == (6, 12)
+
+    # einsum
+    ein = operator.einsum("bij,jk->bik", x, operator.t(x[0]))
+    assert ein.shape == (25, 12, 12)
+
+    # tril
+    triangle = operator.tril(ein, diagonal=-1)
+    assert triangle.shape == (25, 12, 12)
+    assert np.sum([m[0, 0] + m[11, 11] + m[0, 11] for m in triangle]) == 0

oodeel-0.1.1.dist-info/RECORD

@@ -1,46 +0,0 @@
-oodeel/__init__.py,sha256=2q31pS-ZP2CnwNQfZZH3G8iwjYBT2qOwS9Fpl9NERC4,1343
-oodeel/datasets/__init__.py,sha256=BdpBDtWoGS3fgJCRU6TTCcYodzVGYANv5q4Abqhp1RU,1332
-oodeel/datasets/data_handler.py,sha256=kxWKmUjdKKtOOWIPXbt96Bhg8SxVcFUhBCD2XEHx5Ac,7904
-oodeel/datasets/ooddataset.py,sha256=tq3ZIjLNe3iWhjFkmpI4jomLJJ4-CfmsuHoTFR25gBY,13360
-oodeel/datasets/tf_data_handler.py,sha256=FmqWoV05yIQ0lHNBdDE8pTgSdma4jbdjT3YvQOS9KnY,24794
-oodeel/datasets/torch_data_handler.py,sha256=SMa8RvMYp2-f0AI3m5MKeZVFh9G6CMVJl_MNZmiLK74,26814
-oodeel/eval/__init__.py,sha256=lQIUQjczeiRtfIqH2uLNJGubKUN6vPM93mTfY1Qz3bc,1297
-oodeel/eval/metrics.py,sha256=ywU9Y4E_PYckdxllAzUd17T1NtCKq4SfRPigrw78PHI,8589
-oodeel/eval/plots/__init__.py,sha256=YmcFh8RUGvljc-vCew6EiIFMDn0YA_UOfDj4eAv5_Yk,1487
-oodeel/eval/plots/features.py,sha256=tEV1j_1zgqVmC7MOZa7-LyaSQMqL6egH8_FjUwqIhVQ,11815
-oodeel/eval/plots/metrics.py,sha256=3QvLqEB1pAggNHeQJzUpLqL_Ro0MOJ_bPrLcLmf_qpk,4189
-oodeel/eval/plots/plotly.py,sha256=lc22a880TJgUf4H8CSYnXslYRl0I0QqjYsFrf1XjZyA,6034
-oodeel/extractor/__init__.py,sha256=Ew8gLh-xanZsWJe_gKvTa_7HciZ5yTZ-cADKuj7-PCg,1538
-oodeel/extractor/feature_extractor.py,sha256=PE6YczZwRhrd3ZAtEwyWhHl1KmDdvhAOiZHqtYIAnpA,5649
-oodeel/extractor/keras_feature_extractor.py,sha256=IzPDYKS6nHMp-7EIur3mtK5FEVAdPBT_4OTOKTSBRc8,9732
-oodeel/extractor/torch_feature_extractor.py,sha256=VYPoTJfauYkrL2IYmxH9hCT_O8aZKYwhMeB2bZ1wpa4,12465
-oodeel/methods/__init__.py,sha256=TPB-l1dbtvS9YpSG64mFN28Squ6hWNwNMHOFpX5inU0,1556
-oodeel/methods/base.py,sha256=MvjIk7flOGlcufH8UmgWdORJbSIwgPHWUzgBEE8pHxE,11075
-oodeel/methods/dknn.py,sha256=Cbv2y80u8FPho1glb7abC6MGNk2UEjHvEzzMuorBmcc,4264
-oodeel/methods/energy.py,sha256=xcSVs0PibzjFT6pftMuKyZlYkmMdqWJA2WQxw60NHHM,4211
-oodeel/methods/entropy.py,sha256=F5w312f6wpvk9DBHi7LDDcndrMOoPY8OuFCO0ddgOU4,3878
-oodeel/methods/mahalanobis.py,sha256=IIiuyAAvlLkuSZRAmEUGoNANMMuM6KJ-YWb2kc42IsA,7237
-oodeel/methods/mls.py,sha256=BJajmhl3x11Kr-8RwYq59kAn-rveA5ydyoko7ajF_60,4105
-oodeel/methods/odin.py,sha256=E4I2olYH9xvnkgau_2rcPnWL3FFWAh-jxG60-CGxlJA,5254
-oodeel/methods/vim.py,sha256=Z6VeAe_Aaza_rvWL56z1JLVBiRBfmVm-k4ifbrFeH_Q,9485
-oodeel/types/__init__.py,sha256=9TTXjSBfbaDIVMRnclInHI-CBr4L6VZTi61rCJKcTw8,2484
-oodeel/utils/__init__.py,sha256=Lue9BysqeJf5Ej0szafgC9na8IZZR7hWGInJxoEiHUg,1696
-oodeel/utils/general_utils.py,sha256=xc6e7q19ALgMxdCgS7TIyDiMUIGF4ih-aTK1kSlqWoQ,3292
-oodeel/utils/operator.py,sha256=2_jwROZbg0BkZzi6j7SyoGh7iulC1iI-QzZHGZQKHNM,6167
-oodeel/utils/tf_operator.py,sha256=kOp5BvXyY7yf6LdmJg2vRLZHPTFV2HLhlR1na_hZfzI,7212
-oodeel/utils/tf_training_tools.py,sha256=Knih0NGIpBe8G25hV0tpbZi1D2UkFTQCBuQBJYlz3-c,7315
-oodeel/utils/torch_operator.py,sha256=ieiiLlUoexPo5AcFrYG91o2Xe3uzxocWuNce-YOjjHM,7958
-oodeel/utils/torch_training_tools.py,sha256=yOSuLwUw8ENEUUrMJL4PFC35V_XJYmguta2cFjuhpy8,10224
-tests/__init__.py,sha256=lQIUQjczeiRtfIqH2uLNJGubKUN6vPM93mTfY1Qz3bc,1297
-tests/tools_operator.py,sha256=bMCNRUC6uIwVPewTBPrSjPASLxFg6GFzEPLwFobnLIw,5047
-tests/tests_tensorflow/__init__.py,sha256=VuiDSdOBB2jUeobzAW-XTtHJQrcyZpp8i2DyFu0A2RI,1710
-tests/tests_tensorflow/tf_methods_utils.py,sha256=JJqreJaHvhDaGEgw1w7wUhy8YZqYBgkleNC__lYg2Xs,5246
-tests/tests_tensorflow/tools_tf.py,sha256=Z_MEzhwCOF7E7uT-tnfRS3Po9hg0MFzbESP7sMAFBkY,3499
-tests/tests_torch/__init__.py,sha256=3mVxix2Ecn2wUo9DxvzyJBmyOAkv4fVWHHp6uLQHoic,1738
-tests/tests_torch/tools_torch.py,sha256=9ParAhU54AFe4CDYuVo3VGVMypLcMN4Nrk5yz93j-SE,4876
-tests/tests_torch/torch_methods_utils.py,sha256=60dkvArDX3mq9gSato2ghp9BCZeXZ_U64WosCjgdGbE,5234
-oodeel-0.1.1.dist-info/LICENSE,sha256=XrlZ0uYNVeUAF-iEVX21J3CTJjYPgIZUagYSy3Hf0jk,1265
-oodeel-0.1.1.dist-info/METADATA,sha256=WHzwlYu1UvHUd2Qh0REjo-HPGmizFR2fFrzmoROEA9k,17490
-oodeel-0.1.1.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
-oodeel-0.1.1.dist-info/top_level.txt,sha256=zkYRty1FGJ1dkpk-5MU_4uFfBFmcxXoqSwej73xELDs,13
-oodeel-0.1.1.dist-info/RECORD,,