Braindecode 1.3.1.dev178641268__tar.gz → 1.3.1.dev182267549__tar.gz

{braindecode-1.3.1.dev178641268/braindecode.egg-info → braindecode-1.3.1.dev182267549}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: braindecode
-Version: 1.3.1.dev178641268
+Version: 1.3.1.dev182267549
 Summary: Deep learning software to decode EEG, ECG or MEG signals
 Author-email: Robin Tibor Schirrmeister <robintibor@gmail.com>, Bruno Aristimunha Pinto <b.aristimunha@gmail.com>, Alexandre Gramfort <agramfort@meta.com>
 Maintainer-email: Alexandre Gramfort <agramfort@meta.com>, Bruno Aristimunha Pinto <b.aristimunha@gmail.com>, Robin Tibor Schirrmeister <robintibor@gmail.com>

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/braindecode/datasets/base.py

@@ -10,7 +10,6 @@
 
 from __future__ import annotations
 
-import bisect
 import json
 import os
 import shutil
@@ -24,7 +23,7 @@ import mne.io
 import numpy as np
 import pandas as pd
 from mne.utils.docs import deprecated
-from torch.utils.data import ConcatDataset, Dataset, IterableDataset
+from torch.utils.data import ConcatDataset, Dataset
 from typing_extensions import TypeVar
 
 from .bids.hub import HubDatasetMixin
@@ -400,8 +399,6 @@ class BaseConcatDataset(ConcatDataset, HubDatasetMixin, Generic[T]):
         list of RecordDataset
     target_transform : callable | None
         Optional function to call on targets before returning them.
-    lazy : bool, default False
-        If True, defer computing cumulative sizes until length or item access.
     """
 
     datasets: list[T]
@@ -410,12 +407,7 @@ class BaseConcatDataset(ConcatDataset, HubDatasetMixin, Generic[T]):
         self,
         list_of_ds: list[T | BaseConcatDataset[T]],
         target_transform: Callable | None = None,
-        *,
-        lazy: bool = False,
     ):
-        # Adapted from torch.utils.data.ConcatDataset:
-        # added lazy init to defer cumulative size computation until access,
-        # plus _ensure_cumulative_sizes/_get_single_item helpers and __len__ changes.
         # if we get a list of BaseConcatDataset, get all the individual datasets
         flattened_list_of_ds: list[T] = []
         for ds in list_of_ds:
@@ -423,49 +415,14 @@ class BaseConcatDataset(ConcatDataset, HubDatasetMixin, Generic[T]):
                 flattened_list_of_ds.extend(ds.datasets)
             else:
                 flattened_list_of_ds.append(ds)
-
-        # Validate inputs (same validation for both lazy and non-lazy)
-        if len(flattened_list_of_ds) == 0:
-            raise ValueError("datasets should not be an empty iterable")
-        for ds in flattened_list_of_ds:
-            if isinstance(ds, IterableDataset):
-                raise TypeError("ConcatDataset does not support IterableDataset")
-
-        if lazy:
-            Dataset.__init__(self)
-            self.datasets = flattened_list_of_ds
-            # Defer cumulative size computation until first access.
-            self._cumulative_sizes: list[int] | None = None
-        else:
-            super().__init__(flattened_list_of_ds)
+        super().__init__(flattened_list_of_ds)
 
         self.target_transform = target_transform
-        self._lazy = lazy
-
-    @property
-    def cumulative_sizes(self) -> list[int]:
-        """Cumulative sizes of the underlying datasets.
-
-        When the dataset is created with ``lazy=True``, the cumulative sizes
-        are computed on first access and then cached.
-        """
-        return self._ensure_cumulative_sizes()
-
-    @cumulative_sizes.setter
-    def cumulative_sizes(self, value: list[int] | None) -> None:
-        # Keep compatibility with torch.utils.data.ConcatDataset, which assigns
-        # to ``self.cumulative_sizes`` in its __init__.
-        self._cumulative_sizes = value
-
-    def _ensure_cumulative_sizes(self) -> list[int]:
-        if self._cumulative_sizes is None:
-            self._cumulative_sizes = ConcatDataset.cumsum(self.datasets)
-        return self._cumulative_sizes
 
     def _get_sequence(self, indices):
         X, y = list(), list()
         for ind in indices:
-            out_i = self._get_single_item(ind)
+            out_i = super().__getitem__(ind)
             X.append(out_i[0])
             y.append(out_i[1])
 
@@ -474,25 +431,6 @@ class BaseConcatDataset(ConcatDataset, HubDatasetMixin, Generic[T]):
 
         return X, y
 
-    def __len__(self):
-        return self._ensure_cumulative_sizes()[-1]
-
-    def _get_single_item(self, idx: int):
-        if idx < 0:
-            if -idx > len(self):
-                raise ValueError(
-                    "absolute value of index should not exceed dataset length"
-                )
-            idx = len(self) + idx
-
-        cumulative_sizes = self._ensure_cumulative_sizes()
-        dataset_idx = bisect.bisect_right(cumulative_sizes, idx)
-        if dataset_idx == 0:
-            sample_idx = idx
-        else:
-            sample_idx = idx - cumulative_sizes[dataset_idx - 1]
-        return self.datasets[dataset_idx][sample_idx]
-
     def __getitem__(self, idx: int | list):
         """
         ---
@@ -506,7 +444,7 @@ class BaseConcatDataset(ConcatDataset, HubDatasetMixin, Generic[T]):
         if isinstance(idx, Iterable):  # Sample multiple windows
             item = self._get_sequence(idx)
         else:
-            item = self._get_single_item(idx)
+            item = super().__getitem__(idx)
         if self.target_transform is not None:
             item = item[:1] + (self.target_transform(item[1]),) + item[2:]
         return item

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/braindecode/models/__init__.py

@@ -18,7 +18,6 @@ from .eegitnet import EEGITNet
 from .eegminer import EEGMiner
 from .eegnet import EEGNet, EEGNetv4
 from .eegnex import EEGNeX
-from .eegpt import EEGPT
 from .eegsimpleconv import EEGSimpleConv
 from .eegsym import EEGSym
 from .eegtcnet import EEGTCNet
@@ -74,7 +73,6 @@ __all__ = [
     "DeepSleepNet",
     "BrainModule",
     "EEGConformer",
-    "EEGPT",
     "EEGInceptionERP",
     "EEGInceptionMI",
     "EEGITNet",

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/braindecode/models/summary.csv

@@ -8,7 +8,6 @@ CTNet,Motor Imagery,Classification,250,"n_chans, n_outputs, n_times",26900,"CTNe
 Deep4Net,General,Classification,250,"n_chans, n_outputs, n_times",282879,"Deep4Net(n_chans=22, n_outputs=4, n_times=1000)","Convolution"
 DeepSleepNet,Sleep Staging,Classification,256,"n_chans, n_outputs",24744837,"DeepSleepNet(n_chans=1, n_outputs=5, n_times=7680, sfreq=256)","Convolution,Recurrent"
 EEGConformer,General,Classification,250,"n_chans, n_outputs, n_times",789572,"EEGConformer(n_chans=22, n_outputs=4, n_times=1000)","Convolution,Attention/Transformer"
-EEGPT,General,Classification,256,"n_chans, n_outputs, n_times",25477635,"EEGPT(n_chans=58, n_outputs=4, n_times=1024)","Foundation Model,Attention/Transformer"
 EEGInceptionERP,"ERP, SSVEP",Classification,128,"n_chans, n_outputs",14926,"EEGInceptionERP(n_chans=8, n_outputs=2, n_times=128, sfreq=128)","Convolution"
 EEGInceptionMI,Motor Imagery,Classification,250,"n_chans, n_outputs, n_times",558028,"EEGInceptionMI(n_chans=22, n_outputs=4, n_times=1000, n_convs=5, n_filters=12)","Convolution"
 EEGITNet,Motor Imagery,Classification,125,"n_chans, n_outputs, n_times",5212,"EEGITNet(n_chans=22,  n_outputs=4,  n_times=500)","Convolution,Recurrent"

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/braindecode/models/util.py

@@ -73,7 +73,6 @@ models_mandatory_parameters: list[
     ("EEGInceptionMI", ["n_chans", "n_outputs", "n_times", "sfreq"], None),
     ("EEGITNet", ["n_chans", "n_outputs", "n_times"], None),
     ("EEGNet", ["n_chans", "n_outputs", "n_times"], None),
-    ("EEGPT", ["n_chans", "n_outputs", "n_times", "chs_info"], None),
     ("ShallowFBCSPNet", ["n_chans", "n_outputs", "n_times"], None),
     (
         "SleepStagerBlanco2020",

braindecode-1.3.1.dev182267549/braindecode/version.py

@@ -0,0 +1 @@
+__version__ = "1.3.1.dev182267549"

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549/braindecode.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: braindecode
-Version: 1.3.1.dev178641268
+Version: 1.3.1.dev182267549
 Summary: Deep learning software to decode EEG, ECG or MEG signals
 Author-email: Robin Tibor Schirrmeister <robintibor@gmail.com>, Bruno Aristimunha Pinto <b.aristimunha@gmail.com>, Alexandre Gramfort <agramfort@meta.com>
 Maintainer-email: Alexandre Gramfort <agramfort@meta.com>, Bruno Aristimunha Pinto <b.aristimunha@gmail.com>, Robin Tibor Schirrmeister <robintibor@gmail.com>

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/braindecode.egg-info/SOURCES.txt

@@ -70,7 +70,6 @@ braindecode/models/eegitnet.py
 braindecode/models/eegminer.py
 braindecode/models/eegnet.py
 braindecode/models/eegnex.py
-braindecode/models/eegpt.py
 braindecode/models/eegsimpleconv.py
 braindecode/models/eegsym.py
 braindecode/models/eegtcnet.py

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/docs/api.rst

@@ -62,16 +62,12 @@ a common interface for all EEG models and derivate variables names if necessary.
     Some models have pre-trained weights available on the Hugging Face BrainDecode
     organization:
 
-    - :class:`~braindecode.models.BIOT` - Foundation model with pre-trained weights
-    - :class:`~braindecode.models.Labram` - Large Brain Model with pre-trained weights
-    - :class:`~braindecode.models.REVE` - EEG foundation model with pre-trained weights
-    - :class:`~braindecode.models.LUNA` - Universal EEG embedding model with pre-trained
-      weights
-    - :class:`~braindecode.models.BENDR` - Foundation model with pre-trained weights
-    - :class:`~braindecode.models.SignalJEPA` - Self-supervised learning model with
-      pre-trained weights
-    - :class:`~braindecode.models.EEGPT` - Pretrained transformer for universal EEG
-      representation
+    - :class:`BIOT` - Foundation model with pre-trained weights
+    - :class:`Labram` - Large Brain Model with pre-trained weights
+    - :class:`REVE` - EEG foundation model with pre-trained weights
+    - :class:`LUNA` - Universal EEG embedding model with pre-trained weights
+    - :class:`BENDR` - Foundation model with pre-trained weights
+    - :class:`SignalJEPA` - Self-supervised learning model with pre-trained weights
 
     **Example - Loading a pre-trained model:**
 
@@ -142,7 +138,6 @@ a common interface for all EEG models and derivate variables names if necessary.
      EEGITNet
      EEGMiner
      EEGNet
-     EEGPT
      EEGNeX
      EEGSimpleConv
      EEGSym

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/docs/models/models_categorization.rst

@@ -175,7 +175,7 @@ categories can be found in the respective sections below.
 - In contrast, several methods employ **attention/transformer** modules
   (:bdg-info:`Attention/Transformer`) to capture longer-range dependencies efficiently,
   e.g., :class:`EEGConformer`, :class:`CTNet`, :class:`ATCNet`,
-  :class:`AttentionBaseNet`, and :class:`EEGPT`
+  :class:`AttentionBaseNet`
   (:cite:label:`song2022eeg,zhao2024ctnet,altaheri2022atcnet`).
 - **Filterbank-style models** (:bdg-primary:`Filterbank`) explicitly decompose signals
   into multiple bands before (or while) learning, echoing the classic FBCSP pipeline;
@@ -193,8 +193,8 @@ categories can be found in the respective sections below.
 - **Foundation Model / Transformer** (:bdg-danger:`Foundation Model`) approaches
   pretrain attention-based encoders on diverse biosignals and fine-tune for EEG tasks;
   e.g., :class:`BIOT` (:cite:label:`yang2023biot`), :class:`Labram`
-  (:cite:label:`jiang2024large`), and :class:`EEGPT` (:cite:label:`eegpt`). These
-  typically need a heavily self-supervised pre-training before decoding.
+  (:cite:label:`jiang2024large`). These typically need a heavily self-supervised
+  pre-training before decoding.
 - **Graph neural networks** (:bdg-light:`Graph Neural Network`) treat channels/regions
   as nodes with learned (static or dynamic) edges to model functional connectivity
   explicitly; representative EEG-GNN, more common in the epileptic decoding

{braindecode-1.3.1.dev178641268 → braindecode-1.3.1.dev182267549}/docs/whats_new.rst

@@ -27,12 +27,7 @@ Version 1.4  (Source - GitHub)
 
 Enhancements
 ============
-- Add an optional ``lazy`` init path to :class:`braindecode.datasets.base.BaseConcatDataset`
-  to defer cumulative size computation (and any downstream dataset length access) until
-  first access, supporting truly lazy raw loading workflows.
-- Adding :class:`braindecode.models.EEGPT` with HuggingFace Hub integration and
-  pretrained weights available at `braindecode/eegpt-pretrained <https://huggingface.co/braindecode/eegpt-pretrained>`_
-  (:gh:`908` by `Young Truong`_ and `Kuntal Kokate`_)
+- None yet.
 
 API changes
 ============