Revert unwanted merge

.github/workflows/test.yml

@@ -17,7 +17,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, macos-latest]
-        python-version: ['3.9', '3.10', '3.11', '3.12']
+        python-version: ['3.8', '3.9', '3.10', '3.11']
     steps:
       - uses: actions/checkout@v4
       - uses: snok/install-poetry@v1

clinicadl/caps_dataset/data.py

@@ -580,11 +580,11 @@ def _get_mask_paths_and_tensors(
         else:
             for template_ in Template:
                 if preprocessing_.name == template_.name:
-                    template_name = template_.value
+                    template_name = template_
 
             for pattern_ in Pattern:
                 if preprocessing_.name == pattern_.name:
-                    pattern = pattern_.value
+                    pattern = pattern_
 
         mask_location = caps_directory / "masks" / f"tpl-{template_name}"
 

clinicadl/network/cnn/random.py

@@ -208,7 +208,7 @@ def fc_dict_design(n_fcblocks, convolutions, initial_shape, n_classes=2):
         out_channels = last_conv["out_channels"]
         flattened_shape = np.ceil(np.array(initial_shape) / 2**n_conv)
         flattened_shape[0] = out_channels
-        in_features = np.prod(flattened_shape)
+        in_features = np.product(flattened_shape)
 
         # Sample number of FC layers
         ratio = (in_features / n_classes) ** (1 / n_fcblocks)

clinicadl/trainer/tasks_utils.py

@@ -1,20 +1,31 @@
+from abc import abstractmethod
 from typing import Any, Dict, List, Optional, Sequence, Tuple, Type, Union
 
 import numpy as np
 import pandas as pd
 import torch
+import torch.distributed as dist
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    computed_field,
+    model_validator,
+)
 from torch import Tensor, nn
+from torch.amp import autocast
 from torch.nn.functional import softmax
-from torch.utils.data import Sampler, sampler
+from torch.nn.modules.loss import _Loss
+from torch.utils.data import DataLoader, Sampler, sampler
 from torch.utils.data.distributed import DistributedSampler
 
 from clinicadl.caps_dataset.data import CapsDataset
 from clinicadl.metrics.metric_module import MetricModule
+from clinicadl.network.network import Network
 from clinicadl.trainer.config.train import TrainConfig
+from clinicadl.utils import cluster
 from clinicadl.utils.enum import (
     ClassificationLoss,
     ClassificationMetric,
-    Mode,
     ReconstructionLoss,
     ReconstructionMetric,
     RegressionLoss,
@@ -238,7 +249,7 @@ def save_outputs(network_task: Union[str, Task]):
 
 def generate_test_row(
     network_task: Union[str, Task],
-    mode: Mode,
+    mode: str,
     metrics_module,
     n_classes: int,
     idx: int,
@@ -263,7 +274,7 @@ def generate_test_row(
             [
                 data["participant_id"][idx],
                 data["session_id"][idx],
-                data[f"{mode.value}_id"][idx].item(),
+                data[f"{mode}_id"][idx].item(),
                 data["label"][idx].item(),
                 prediction,
             ]
@@ -275,7 +286,7 @@ def generate_test_row(
             [
                 data["participant_id"][idx],
                 data["session_id"][idx],
-                data[f"{mode.value}_id"][idx].item(),
+                data[f"{mode}_id"][idx].item(),
                 data["label"][idx].item(),
                 outputs[idx].item(),
             ]
@@ -287,7 +298,7 @@ def generate_test_row(
         row = [
             data["participant_id"][idx],
             data["session_id"][idx],
-            data[f"{mode.value}_id"][idx].item(),
+            data[f"{mode}_id"][idx].item(),
         ]
 
         for metric in evaluation_metrics(Task.RECONSTRUCTION):

docs/Installation.md

@@ -30,7 +30,7 @@ bash /tmp/miniconda-installer.sh
 The latest release of ClinicaDL can be installed using `pip` as follows:
 
 ```{.sourceCode .bash}
-conda create --name clinicadlEnv python=3.11
+conda create --name clinicadlEnv python=3.8
 conda activate clinicadlEnv
 pip install clinicadl
 ```

environment.yml

@@ -3,4 +3,4 @@ channels:
   - defaults
   - conda-forge
 dependencies:
-  - python=3.11
+  - python=3.9

pyproject.toml

@@ -27,15 +27,15 @@ classifiers = [
 ]
 
 [tool.poetry.dependencies]
-python = ">=3.9,<3.13"
-torch = "^2.3.0"
+python = ">=3.8,<3.12"
+torch = "^2.1.0"
 torchvision = "*"
 tensorboard = "*"
 toml = "*"
-pandas = "^2"
-numpy = "^2"
-scikit-learn = "^1"
-scikit-image = "^0.24"
+pandas = "^1.2"
+numpy = "^1.17"
+scikit-learn = "^1.0"
+scikit-image = "^0.21"
 joblib = "^1.2.0"
 click = "^8"
 click-option-group = "^0.5"