iapytoo 0.0.4.dev3__tar.gz → 0.0.6__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (80) hide show
  1. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/.gitignore +4 -1
  2. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/PKG-INFO +3 -2
  3. iapytoo-0.0.6/dockers/docker-compose.yml +15 -0
  4. iapytoo-0.0.6/dockers/requirements.txt +6 -0
  5. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/dockers/run-python.sh +1 -2
  6. iapytoo-0.0.6/examples/config_infer.yml +17 -0
  7. iapytoo-0.0.6/examples/config_mnist.yml +26 -0
  8. iapytoo-0.0.6/examples/config_wgan.yml +29 -0
  9. iapytoo-0.0.6/examples/mlflow_infer.py +37 -0
  10. iapytoo-0.0.6/examples/mnist.py +186 -0
  11. iapytoo-0.0.6/examples/mnist_infer.py +41 -0
  12. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/examples/wgan_mnist.py +26 -29
  13. iapytoo-0.0.6/examples/zipped_runs/selected_runs.zip +0 -0
  14. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/_version.py +9 -4
  15. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/dataset/__init__.py +4 -8
  16. iapytoo-0.0.6/iapytoo/metrics/collection.py +39 -0
  17. iapytoo-0.0.6/iapytoo/metrics/metric.py +46 -0
  18. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/metrics/predefined.py +13 -12
  19. iapytoo-0.0.6/iapytoo/predictions/__init__.py +69 -0
  20. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/predictions/feature.py +1 -2
  21. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/predictions/plotters.py +39 -11
  22. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/predictions/predictors.py +0 -2
  23. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/predictions/types.py +0 -2
  24. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/checkpoint.py +18 -9
  25. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/factories.py +188 -112
  26. iapytoo-0.0.6/iapytoo/train/inference.py +175 -0
  27. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/logger.py +52 -54
  28. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/loss.py +4 -5
  29. iapytoo-0.0.6/iapytoo/train/mlflow_model.py +88 -0
  30. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/training.py +61 -101
  31. iapytoo-0.0.6/iapytoo/train/valuator.py +61 -0
  32. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/wgan.py +104 -46
  33. iapytoo-0.0.6/iapytoo/utils/__init__.py +0 -0
  34. iapytoo-0.0.6/iapytoo/utils/arguments.py +30 -0
  35. iapytoo-0.0.6/iapytoo/utils/cf/config.yml +21 -0
  36. iapytoo-0.0.6/iapytoo/utils/config.py +344 -0
  37. iapytoo-0.0.6/iapytoo/utils/model_config.py +74 -0
  38. iapytoo-0.0.6/iapytoo/utils/singleton.py +15 -0
  39. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo.egg-info/PKG-INFO +3 -2
  40. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo.egg-info/SOURCES.txt +19 -3
  41. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo.egg-info/top_level.txt +2 -0
  42. iapytoo-0.0.6/mlflow/relocate_mlruns.py +67 -0
  43. iapytoo-0.0.6/mlflow/zip_selected_runs.py +80 -0
  44. iapytoo-0.0.6/tests/test_config.py +100 -0
  45. iapytoo-0.0.6/tests/test_factories.py +126 -0
  46. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/tests/test_metrics.py +29 -8
  47. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/tests/test_model.py +3 -3
  48. iapytoo-0.0.4.dev3/examples/config_mnist.json +0 -19
  49. iapytoo-0.0.4.dev3/examples/config_wgan.json +0 -25
  50. iapytoo-0.0.4.dev3/examples/mnist.py +0 -106
  51. iapytoo-0.0.4.dev3/iapytoo/metrics/__init__.py +0 -82
  52. iapytoo-0.0.4.dev3/iapytoo/metrics/metric.py +0 -39
  53. iapytoo-0.0.4.dev3/iapytoo/predictions/__init__.py +0 -78
  54. iapytoo-0.0.4.dev3/iapytoo/utils/arguments.py +0 -45
  55. iapytoo-0.0.4.dev3/iapytoo/utils/config.py +0 -123
  56. iapytoo-0.0.4.dev3/iapytoo/utils/meta_singleton.py +0 -17
  57. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/.github/workflows/python-app.yml +0 -0
  58. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/LICENSE +0 -0
  59. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/MANIFEST.in +0 -0
  60. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/README.md +0 -0
  61. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/dockers/Dockerfile +0 -0
  62. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/dockers/Readme.md +0 -0
  63. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/__init__.py +0 -0
  64. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/dataset/scaling.py +0 -0
  65. {iapytoo-0.0.4.dev3/iapytoo/train → iapytoo-0.0.6/iapytoo/metrics}/__init__.py +0 -0
  66. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/metrics/distrib.py +0 -0
  67. {iapytoo-0.0.4.dev3/iapytoo/utils → iapytoo-0.0.6/iapytoo/train}/__init__.py +0 -0
  68. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/train/context.py +0 -0
  69. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/cf/config.json +0 -0
  70. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/cf/config_test.json +0 -0
  71. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/display.py +0 -0
  72. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/iterative_mean.py +0 -0
  73. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/time_utils.py +0 -0
  74. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo/utils/timer.py +0 -0
  75. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/iapytoo.egg-info/dependency_links.txt +0 -0
  76. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/pyproject.toml +0 -0
  77. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/requirements.txt +0 -0
  78. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/setup.cfg +0 -0
  79. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/tests/test_dataset.py +0 -0
  80. {iapytoo-0.0.4.dev3 → iapytoo-0.0.6}/tests/test_scaling.py +0 -0
@@ -165,4 +165,7 @@ cython_debug/
165
165
 
166
166
 
167
167
  mlruns/
168
- data/
168
+ data/
169
+ dockers/Dockerfile.ifpen
170
+ iapytoo/_version.py
171
+
@@ -1,12 +1,13 @@
1
- Metadata-Version: 2.2
1
+ Metadata-Version: 2.4
2
2
  Name: iapytoo
3
- Version: 0.0.4.dev3
3
+ Version: 0.0.6
4
4
  Summary: Un toolkit pour vite coder un modèle pytorch d'IA utilisant MLflow
5
5
  Author-email: Jean-François Lecomte <zwergon@free.fr>
6
6
  License: Apache-2.0
7
7
  Requires-Python: >=3.7
8
8
  Description-Content-Type: text/markdown
9
9
  License-File: LICENSE
10
+ Dynamic: license-file
10
11
 
11
12
  [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
12
13
  [![Flake8](
@@ -0,0 +1,15 @@
1
+ version: "3.8"
2
+
3
+ services:
4
+ iapytoo:
5
+ build:
6
+ context: .
7
+ dockerfile: Dockerfile.ifpen
8
+ args:
9
+ http_proxy: ${http_proxy}
10
+ https_proxy: ${https_proxy}
11
+ no_proxy: ${no_proxy}
12
+ image: iapytoo:latest
13
+
14
+ env_file:
15
+ - .env
@@ -0,0 +1,6 @@
1
+ tqdm
2
+ numpy
3
+ matplotlib
4
+ mlflow
5
+ seaborn
6
+ iapytoo
@@ -5,5 +5,4 @@ docker run --rm --gpus all \
5
5
  -w /app \
6
6
  -e PYTHONPATH=/app \
7
7
  iapytoo:latest \
8
- python3 examples/mnist.py -c examples/config_mnist.json data
9
-
8
+ python3 examples/mnist.py --yaml examples/config_mnist.yml
@@ -0,0 +1,17 @@
1
+ project: "iapytoo"
2
+ run: "infer"
3
+ tracking_uri: null
4
+
5
+ dataset:
6
+ path: "../data"
7
+ n_steps_by_batch: 10
8
+ num_workers: 2
9
+ batch_size: 64
10
+
11
+ metrics:
12
+ names: ['accuracy', 'r2']
13
+
14
+ model:
15
+ type: "mlflow"
16
+ predictor: "max"
17
+ run_id: "5bfea07ccb91434fafe6fc1def3c2011"
@@ -0,0 +1,26 @@
1
+ project: "iapytoo"
2
+ run: "run"
3
+ tracking_uri: null
4
+
5
+ dataset:
6
+ path: "../data"
7
+ n_steps_by_batch: 10
8
+ num_workers: 2
9
+ batch_size: 64
10
+
11
+ training:
12
+ learning_rate: 1e-4
13
+ loss: "nll"
14
+ optimizer: "adam"
15
+ scheduler: "step"
16
+ epochs: 2
17
+ #tqdm: false
18
+
19
+
20
+ metrics:
21
+ names: ['accuracy']
22
+
23
+ model:
24
+ model: "mnist"
25
+ predictor: "max"
26
+ kernel_size: 3
@@ -0,0 +1,29 @@
1
+ project: wgan
2
+ run: wgan
3
+ tracking_uri: null
4
+
5
+ dataset:
6
+ path: "./data"
7
+ n_steps_by_batch: 10
8
+ num_workers: 2
9
+ batch_size: 256
10
+
11
+ training:
12
+ learning_rate: 0.0002
13
+ optimizer: adam
14
+ scheduler: step
15
+ gamma: 0.7
16
+ # betas:
17
+ # - 0.5
18
+ # - 0.99
19
+ epochs: 11
20
+ loss: "wasserstein"
21
+
22
+ model:
23
+ type: "gan"
24
+ valuator: "gan"
25
+ generator: generator
26
+ discriminator: critic
27
+ n_critic: 1
28
+ lambda_gp: 10
29
+ noise_dim: 100
@@ -0,0 +1,37 @@
1
+ from torchvision import datasets
2
+ import mlflow
3
+ import numpy as np
4
+ from PIL import Image
5
+
6
+
7
+ def get_model_input(idx=0):
8
+
9
+ dataset1 = datasets.MNIST(
10
+ "../data",
11
+ train=True,
12
+ download=False
13
+ )
14
+ # prend une image du dataset MNIST de train
15
+ pic: Image = dataset1[idx][0]
16
+ expected = dataset1[idx][1]
17
+ # cree le numpy array
18
+ img_input = np.array(pic.getdata()).reshape(
19
+ 1, pic.size[0], pic.size[1]).astype(np.float64)
20
+ img_input /= 255.
21
+
22
+ # crée un batch avec une image [1, 1, 28, 28]
23
+ return np.array([img_input]), expected
24
+
25
+
26
+ if __name__ == "__main__":
27
+ import mlflow.pyfunc as mp
28
+ run_id = "5bfea07ccb91434fafe6fc1def3c2011"
29
+
30
+ logged_model = f'runs:/{run_id}/model'
31
+
32
+ # Load model as a PyFuncModel.
33
+ loaded_model: mp.PyFuncModel = mlflow.pyfunc.load_model(logged_model)
34
+
35
+ model_input, expected = get_model_input()
36
+ predicted = loaded_model.predict(model_input)
37
+ print(predicted[0], expected)
@@ -0,0 +1,186 @@
1
+ import numpy as np
2
+ import torch
3
+ import torch.nn as nn
4
+ import torch.nn.functional as F
5
+ from torch.utils.data import DataLoader
6
+ from torchvision import datasets, transforms
7
+ from torch.optim.lr_scheduler import LambdaLR
8
+ from PIL import Image
9
+
10
+
11
+ from iapytoo.predictions.plotters import ConfusionPlotter
12
+ from iapytoo.train.factories import Model, Scheduler, Factory
13
+ from iapytoo.utils.config import ConfigFactory, Config
14
+ from iapytoo.train.training import Training
15
+ from iapytoo.train.mlflow_model import MlflowTransform, IMlfowModelProvider
16
+
17
+
18
+ from mlflow.types.schema import TensorSpec, Schema
19
+ from mlflow.models import ModelSignature
20
+
21
+
22
+ class MnistModel(Model):
23
+ def __init__(self, loader, config: Config):
24
+ super(MnistModel, self).__init__(loader, config)
25
+ self.conv1 = nn.Conv2d(1, 32, 3, 1)
26
+ self.conv2 = nn.Conv2d(32, 64, 3, 1)
27
+ self.dropout1 = nn.Dropout(0.25)
28
+ self.dropout2 = nn.Dropout(0.5)
29
+ self.fc1 = nn.Linear(9216, 128)
30
+ self.fc2 = nn.Linear(128, 10)
31
+
32
+ def forward(self, x):
33
+ x = self.conv1(x)
34
+ x = F.relu(x)
35
+ x = self.conv2(x)
36
+ x = F.relu(x)
37
+ x = F.max_pool2d(x, 2)
38
+ x = self.dropout1(x)
39
+ x = torch.flatten(x, 1)
40
+ x = self.fc1(x)
41
+ x = F.relu(x)
42
+ x = self.dropout2(x)
43
+ x = self.fc2(x)
44
+ output = F.log_softmax(x, dim=1)
45
+ return output
46
+
47
+ def predict(self, model_output):
48
+ return torch.argmax(model_output, dim=1)
49
+
50
+
51
+ class MnistScheduler(Scheduler):
52
+ def __init__(self, optimizer, config: Config) -> None:
53
+ super().__init__(optimizer, config)
54
+
55
+ def lr_lambda(epoch):
56
+ # LR to be 0.1 * (1/1+0.01*epoch)
57
+ return 0.995**epoch
58
+
59
+ self.lr_scheduler = LambdaLR(optimizer, lr_lambda=lr_lambda)
60
+
61
+
62
+ class MnistTransform(MlflowTransform):
63
+
64
+ def __init__(self) -> None:
65
+ super().__init__(
66
+ transform=transforms.Compose(
67
+ [
68
+ transforms.ToTensor(),
69
+ transforms.Normalize((0.1307,), (0.3081,))
70
+ ]
71
+ )
72
+ )
73
+
74
+ # override
75
+ def __call__(self, model_input, *args, **kwds):
76
+
77
+ images_pil = []
78
+ for i in range(model_input.shape[0]):
79
+ # Extraire l'image (en supprimant la dimension du canal si nécessaire)
80
+ # Supposons que l'image est en niveaux de gris
81
+ img_array = model_input[i, 0, :, :]
82
+
83
+ # Normaliser l'image si nécessaire (par exemple, si les valeurs sont entre 0 et 1)
84
+ img_array = (img_array * 255).astype(np.uint8)
85
+
86
+ # Créer une image PIL
87
+ # 'L' pour les niveaux de gris
88
+ img_pil = Image.fromarray(img_array, 'L')
89
+ images_pil.append(img_pil)
90
+
91
+ model_input_tensor = torch.stack(
92
+ [self.transform(img) for img in images_pil])
93
+
94
+ return model_input_tensor
95
+
96
+
97
+ class MnistMlfowModel(IMlfowModelProvider):
98
+
99
+ def __init__(self):
100
+ X = np.random.rand(1, 1, 28, 28)
101
+ Y = np.zeros(shape=(1,), dtype=np.int64)
102
+ # add batch size
103
+ x_shape = list(X.shape)
104
+ x_shape[0] = -1
105
+ y_shape = list(Y.shape)
106
+ y_shape[0] = -1
107
+
108
+ input_schema = Schema(
109
+ [TensorSpec(type=X.dtype, shape=x_shape)])
110
+ output_schema = Schema(
111
+ [TensorSpec(type=Y.dtype, shape=y_shape)])
112
+
113
+ self.signature = ModelSignature(
114
+ inputs=input_schema, outputs=output_schema)
115
+ self.input_example = X
116
+ self.transform: MlflowTransform = MnistTransform()
117
+
118
+ # override
119
+ def get_transform(self) -> MlflowTransform:
120
+ return self.transform
121
+
122
+ # override
123
+ def get_signature(self) -> ModelSignature:
124
+ return self.signature
125
+
126
+ # override
127
+ def get_input_example(self) -> np.array:
128
+ return self.input_example
129
+
130
+
131
+ class MnistTraining(Training):
132
+
133
+ def __init__(self, config: Config) -> None:
134
+ super().__init__(config)
135
+ self.predictions.add_plotter(ConfusionPlotter())
136
+ self.mlflow_model_provider: IMlfowModelProvider = MnistMlfowModel()
137
+
138
+
139
+ if __name__ == "__main__":
140
+ from iapytoo.utils.arguments import parse_args
141
+
142
+ factory = Factory()
143
+ factory.register_model("mnist", MnistModel)
144
+ factory.register_scheduler("mnist", MnistScheduler)
145
+
146
+ args = parse_args()
147
+
148
+ if args.run_id is not None:
149
+ config = ConfigFactory.from_run_id(args.run_id, args.tracking_uri)
150
+ config.training.epochs = args.epochs
151
+ else:
152
+ # INPUT Parameters
153
+ config = ConfigFactory.from_yaml(args.yaml)
154
+
155
+ Training.seed(config)
156
+
157
+ training = MnistTraining(config)
158
+ mflow_transform: MlflowTransform = training.get_transform()
159
+ train_dataset = datasets.MNIST(
160
+ config.dataset.path,
161
+ train=True,
162
+ download=True,
163
+ transform=mflow_transform.transform
164
+ )
165
+
166
+ test_dataset = datasets.MNIST(
167
+ config.dataset.path,
168
+ train=False,
169
+ transform=mflow_transform.transform
170
+ )
171
+
172
+ train_loader = DataLoader(
173
+ train_dataset,
174
+ batch_size=config.dataset.batch_size,
175
+ num_workers=config.dataset.num_workers
176
+ )
177
+ test_loader = DataLoader(
178
+ test_dataset, batch_size=config.dataset.batch_size,
179
+ num_workers=config.dataset.num_workers
180
+ )
181
+
182
+ training.fit(
183
+ train_loader=train_loader,
184
+ valid_loader=test_loader,
185
+ run_id=args.run_id
186
+ )
@@ -0,0 +1,41 @@
1
+ from torchvision import datasets
2
+ from torch.utils.data import DataLoader
3
+ from iapytoo.predictions.plotters import ConfusionPlotter
4
+ from iapytoo.utils.config import Config, ConfigFactory, DatasetConfig
5
+
6
+ from iapytoo.train.inference import MLFlowInference
7
+ from iapytoo.train.mlflow_model import MlflowTransform
8
+
9
+
10
+ class MnistInference(MLFlowInference):
11
+ def __init__(self, config: Config) -> None:
12
+ super().__init__(config)
13
+ self.predictions.add_plotter(ConfusionPlotter())
14
+
15
+
16
+ if __name__ == "__main__":
17
+ import os
18
+
19
+ # INPUT Parameters
20
+ config = ConfigFactory.from_yaml(os.path.join(
21
+ os.path.dirname(__file__), "config_infer.yml"))
22
+
23
+ inference = MnistInference(config)
24
+
25
+ transform: MlflowTransform = inference.get_transform()
26
+
27
+ dataset_config: DatasetConfig = config.dataset
28
+
29
+ test_dataset = datasets.MNIST(
30
+ dataset_config.path,
31
+ train=False,
32
+ transform=transform.transform
33
+ )
34
+
35
+ test_loader = DataLoader(
36
+ test_dataset,
37
+ batch_size=dataset_config.batch_size,
38
+ num_workers=2
39
+ )
40
+
41
+ inference.predict(loader=test_loader)
@@ -1,25 +1,16 @@
1
1
  import os
2
- import torch
3
- import torch.nn as nn
4
- import torch.optim as optim
5
- import numpy as np
6
- from torch.autograd import grad
7
-
8
- from iapytoo.train.wgan import WGAN
9
- from iapytoo.utils.config import Config
10
-
11
-
12
- from iapytoo.train.factories import Model, ModelFactory, OptimizerFactory
13
- from iapytoo.predictions.plotters import PredictionPlotter, Fake2DPlotter
14
2
 
15
3
  import torch
16
4
  import torch.nn as nn
17
5
  from torch.utils.data import Dataset
18
- import torchvision
19
- import numpy as np
20
6
  from torchvision.datasets import MNIST
21
7
  import torchvision.transforms as transforms
22
- import matplotlib.pyplot as plt
8
+
9
+
10
+ from iapytoo.train.wgan import WGAN
11
+ from iapytoo.utils.config import ConfigFactory, Config
12
+ from iapytoo.train.factories import Model, Factory
13
+ from iapytoo.predictions.plotters import Fake2DPlotter
23
14
  from iapytoo.train.factories import WeightInitiator
24
15
 
25
16
 
@@ -47,9 +38,9 @@ class Generator(Model):
47
38
  """
48
39
  return torch.randn(n_samples, z_dim, device=device)
49
40
 
50
- def __init__(self, loader, config):
41
+ def __init__(self, loader, config: Config):
51
42
  super(Generator, self).__init__(loader, config)
52
- self.z_dim = config["noise_dim"]
43
+ self.z_dim = config.model.noise_dim
53
44
  # Filters [1024, 512, 256]
54
45
  # Input_dim = 100
55
46
  # Output_dim = 1 (number of channels)
@@ -107,15 +98,18 @@ class Generator(Model):
107
98
 
108
99
 
109
100
  class Discriminator(Model):
110
- def __init__(self, loader, config):
101
+ def __init__(self, loader, config: Config):
111
102
  super(Discriminator, self).__init__(loader, config)
112
103
  # Filters [256, 512, 1024]
113
104
  # Input_dim = 1 (Cx64x64)
114
105
  # Output_dim = 1
115
106
  self.main_module = nn.Sequential(
116
- # Omitting batch normalization in critic because our new penalized training objective (WGAN with gradient penalty) is no longer valid
117
- # in this setting, since we penalize the norm of the critic's gradient with respect to each input independently and not the enitre batch.
118
- # There is not good & fast implementation of layer normalization --> using per instance normalization nn.InstanceNorm2d()
107
+ # Omitting batch normalization in critic because our new penalized training objective
108
+ # (WGAN with gradient penalty) is no longer valid
109
+ # in this setting, since we penalize the norm of the critic's gradient with respect
110
+ # to each input independently and not the enitre batch.
111
+ # There is not good & fast implementation of layer normalization -->
112
+ # using per instance normalization nn.InstanceNorm2d()
119
113
  # Image (Cx32x32)
120
114
  nn.Conv2d(
121
115
  in_channels=1, out_channels=256, kernel_size=4, stride=2, padding=1
@@ -167,7 +161,8 @@ class LatentDataset(Dataset):
167
161
 
168
162
 
169
163
  if __name__ == "__main__":
170
- config = Config(os.path.join(os.path.dirname(__file__), "config_wgan.json"))
164
+ config = ConfigFactory.from_yaml(os.path.join(
165
+ os.path.dirname(__file__), "config_wgan.yml"))
171
166
 
172
167
  transform = transforms.Compose(
173
168
  [
@@ -178,22 +173,24 @@ if __name__ == "__main__":
178
173
  )
179
174
 
180
175
  # load training data
181
- trainset = MNIST(config.data_dir, download=True, train=True, transform=transform)
176
+ trainset = MNIST(config.dataset.path, download=True,
177
+ train=True, transform=transform)
182
178
 
183
179
  trainloader = torch.utils.data.DataLoader(
184
- trainset, batch_size=config.batch_size, shuffle=True, drop_last=True
180
+ trainset, batch_size=config.dataset.batch_size, shuffle=True, drop_last=True
185
181
  )
186
182
 
187
- latentset = LatentDataset(config.noise_dim, size=16)
183
+ latentset = LatentDataset(config.model.noise_dim, size=16)
188
184
 
189
- valid_loader = torch.utils.data.DataLoader(latentset, batch_size=1, shuffle=False)
185
+ valid_loader = torch.utils.data.DataLoader(
186
+ latentset, batch_size=1, shuffle=False)
190
187
 
191
188
  item = next(iter(valid_loader))
192
189
  print(item.shape)
193
190
 
194
- model_factory = ModelFactory()
195
- model_factory.register_model("generator", Generator)
196
- model_factory.register_model("critic", Discriminator)
191
+ factory = Factory()
192
+ factory.register_model("generator", Generator)
193
+ factory.register_model("critic", Discriminator)
197
194
 
198
195
  wgan = WGAN(config)
199
196
  wgan.predictions.add_plotter(Fake2DPlotter())
@@ -1,8 +1,13 @@
1
- # file generated by setuptools_scm
1
+ # file generated by setuptools-scm
2
2
  # don't change, don't track in version control
3
+
4
+ __all__ = ["__version__", "__version_tuple__", "version", "version_tuple"]
5
+
3
6
  TYPE_CHECKING = False
4
7
  if TYPE_CHECKING:
5
- from typing import Tuple, Union
8
+ from typing import Tuple
9
+ from typing import Union
10
+
6
11
  VERSION_TUPLE = Tuple[Union[int, str], ...]
7
12
  else:
8
13
  VERSION_TUPLE = object
@@ -12,5 +17,5 @@ __version__: str
12
17
  __version_tuple__: VERSION_TUPLE
13
18
  version_tuple: VERSION_TUPLE
14
19
 
15
- __version__ = version = '0.0.4.dev3'
16
- __version_tuple__ = version_tuple = (0, 0, 4, 'dev3')
20
+ __version__ = version = '0.0.6'
21
+ __version_tuple__ = version_tuple = (0, 0, 6)
@@ -13,8 +13,7 @@ class DummyVisionDataset(VisionDataset):
13
13
  np.random.seed(12)
14
14
  A = np.array(range(self.dim * self.dim)).reshape((self.dim, self.dim))
15
15
  self.images = np.zeros(
16
- shape=(self.length, self.n_channels,
17
- self.dim, self.dim), dtype=np.float32
16
+ shape=(self.length, self.n_channels, self.dim, self.dim), dtype=np.float32
18
17
  )
19
18
  for idx in range(self.length):
20
19
  for c in range(self.n_channels):
@@ -24,10 +23,8 @@ class DummyVisionDataset(VisionDataset):
24
23
 
25
24
  self.targets = np.vstack(
26
25
  [
27
- np.array(range(self.length)) +
28
- np.random.normal(size=(self.length)),
29
- np.array(range(self.length)) +
30
- np.random.normal(size=(self.length)),
26
+ np.array(range(self.length)) + np.random.normal(size=(self.length)),
27
+ np.array(range(self.length)) + np.random.normal(size=(self.length)),
31
28
  ]
32
29
  )
33
30
 
@@ -49,8 +46,7 @@ class DummyLabelDataset(VisionDataset):
49
46
  np.random.seed(12)
50
47
  A = np.array(range(self.dim * self.dim)).reshape((self.dim, self.dim))
51
48
  self.images = np.zeros(
52
- shape=(self.length, self.n_channels,
53
- self.dim, self.dim), dtype=np.float32
49
+ shape=(self.length, self.n_channels, self.dim, self.dim), dtype=np.float32
54
50
  )
55
51
  for idx in range(self.length):
56
52
  for c in range(self.n_channels):
@@ -0,0 +1,39 @@
1
+ from iapytoo.utils.config import Config
2
+ from iapytoo.metrics.metric import Metric
3
+
4
+
5
+ class MetricsCollection(Metric):
6
+ def __init__(self, tag: str, metric_names: list[str], config: Config):
7
+ from iapytoo.train.factories import Factory, MetricError
8
+ super().__init__(tag, config)
9
+ self.metrics = {}
10
+ try:
11
+ factory = Factory()
12
+ for n in metric_names:
13
+ self.metrics[f"{tag}_{n}"] = factory.create_metric(n, config)
14
+ except MetricError as er:
15
+ print(f"Unable to create metric : {er}")
16
+
17
+ def to(self, device):
18
+ for m in self.metrics.values():
19
+ m.to(device)
20
+
21
+ # overwrite
22
+ def update(self, Y_hat, Y):
23
+ for m in self.metrics.values():
24
+ m.update(Y_hat, Y)
25
+
26
+ # overwrite
27
+ def compute(self):
28
+ self.results = {}
29
+ for k, m in self.metrics.items():
30
+ results = m.compute()
31
+ for name, result in results.items():
32
+ # prepend the name of the collection to all inner results : validation_accuracy for example
33
+ self.results[f"{self.name}_{name}"] = result
34
+ return self.results
35
+
36
+ # overwrite
37
+ def reset(self) -> None:
38
+ for m in self.metrics.values():
39
+ m.reset()
@@ -0,0 +1,46 @@
1
+ import torch
2
+
3
+ from iapytoo.utils.config import Config
4
+
5
+
6
+ class Metric:
7
+ def __init__(self, name, config: Config, with_target=True):
8
+ from iapytoo.train.factories import Factory
9
+ self.with_target = with_target
10
+ self.name = name
11
+ self.config = config
12
+ self.predictor = Factory().create_predictor(config)
13
+
14
+ # First dimension is for the concatenation.
15
+ self.outputs = torch.zeros(size=(0,))
16
+ if self.with_target:
17
+ self.target = torch.zeros(size=(0,))
18
+ self.results = {}
19
+
20
+ @property
21
+ def device(self):
22
+ return self.outputs.device
23
+
24
+ @property
25
+ def predicted(self):
26
+ return self.predictor(self.outputs)
27
+
28
+ def to(self, device):
29
+ self.outputs = self.outputs.to(device)
30
+ if self.with_target:
31
+ self.target = self.target.to(device)
32
+ return self
33
+
34
+ def update(self, outputs, Y):
35
+ self.outputs = torch.cat((self.outputs, outputs), dim=0)
36
+ if self.with_target:
37
+ self.target = torch.cat((self.target, Y), dim=0)
38
+
39
+ def reset(self):
40
+ device = self.device
41
+ self.outputs = torch.zeros(size=(0,), device=device)
42
+ if self.with_target:
43
+ self.target = torch.zeros(size=(0,), device=self.device)
44
+
45
+ def compute(self):
46
+ pass