From 89445fe86fd45675acb20ab676b7530475d0fcbb Mon Sep 17 00:00:00 2001
From: Aske-Rosted <askerosted@gmail.com>
Date: Fri, 19 May 2023 17:21:08 +0900
Subject: [PATCH 1/3] The implementation of the RNN_DynEdge model
---
.../models/components/DOM_handling.py | 147 ++++++++++++++++++
src/graphnet/models/gnn/__init__.py | 1 +
src/graphnet/models/gnn/node_rnn.py | 81 ++++++++++
src/graphnet/models/gnn/rnn_dynedge.py | 90 +++++++++++
4 files changed, 319 insertions(+)
create mode 100644 src/graphnet/models/components/DOM_handling.py
create mode 100644 src/graphnet/models/gnn/node_rnn.py
create mode 100644 src/graphnet/models/gnn/rnn_dynedge.py
diff --git a/src/graphnet/models/components/DOM_handling.py b/src/graphnet/models/components/DOM_handling.py
new file mode 100644
index 000000000..0bfc95b4d
--- /dev/null
+++ b/src/graphnet/models/components/DOM_handling.py
@@ -0,0 +1,147 @@
+"""Functions for handling event level DOM data."""
+from typing import Tuple, Optional
+from torch_geometric.nn.pool import knn_graph
+
+import torch
+
+
+@torch.jit.script
+def append_dom_id(
+ tensor: torch.Tensor,
+ batch: torch.Tensor,
+ device: torch.device = torch.device("cpu"),
+) -> torch.Tensor:
+ """Assign a unique ID to each DOM.
+
+ The ID is assigned based on the position of the DOM in the input tensor. requires x,y,z as the first three columns of the input tensor.
+
+ Args:
+ tensor (torch.Tensor): Input tensor.
+ batch (torch.Tensor, optional): Batch tensor. Defaults to None.
+ device (torch.device): Device to use.
+
+ Returns:
+ torch.Tensor: Tensor of DOM IDs.
+ """
+ inverse_matrix = torch.zeros(
+ (tensor.shape[0], 3), device=device, dtype=torch.int64
+ )
+ for i in range(3):
+ _, inverse = torch.unique(tensor[:, i], return_inverse=True)
+ inverse_matrix[:, i] = inverse
+
+ if batch is None:
+ batch = torch.zeros(tensor.shape[0], device=device, dtype=torch.int64)
+
+ inverse_matrix = torch.hstack([batch.unsqueeze(1), inverse_matrix])
+ for i in range(3):
+ inverse_matrix[:, 1] = inverse_matrix[:, 0] + (
+ (torch.max(inverse_matrix[:, 0]) + 1) * (inverse_matrix[:, 1] + 1)
+ )
+
+ _, inverse_matrix[:, 1] = torch.unique(
+ inverse_matrix[:, 1], return_inverse=True
+ )
+
+ inverse_matrix = inverse_matrix[:, -(3 - i) :]
+
+ inverse_matrix = inverse_matrix.flatten()
+ tensor = torch.hstack([tensor, inverse_matrix.unsqueeze(1)])
+ return tensor
+
+
+torch.jit.script
+
+
+def DOM_to_time_series(
+ tensor: torch.Tensor, batch: torch.tensor
+) -> Tuple[torch.Tensor, torch.Tensor]:
+ """Create time series for each activated DOM from dom activation data.
+
+ Returns a time series view of the DOM activations as well as an updated batch tensor. REQURES the DOM ID to be the last column of the input tensor.
+ Args:
+ tensor (torch.Tensor): Input tensor.
+ batch (torch.Tensor): Batch tensor.
+
+ Returns:
+ tensor (torch.Tensor): Times series DOM data.
+ sort_batch (torch.Tensor): Batch tensor.
+ """
+ dom_activation_sort = tensor[:, -1].sort()[-1]
+ tensor, batch = tensor[dom_activation_sort], batch[dom_activation_sort]
+ bin_count = torch.bincount(tensor[:, -1].type(torch.int64)).cumsum(0)
+ batch = batch[bin_count - 1]
+ tensor = tensor[:, :-1]
+ tensor = torch.tensor_split(tensor, bin_count.cpu()[:-1])
+ lengths_index = (
+ torch.as_tensor([v.size(0) for v in tensor]).sort()[-1].flip(0)
+ )
+ batch = batch[lengths_index]
+
+ return tensor, batch
+
+
+torch.jit.script
+
+
+def DOM_time_series_to_pack_sequence(
+ tensor: torch.Tensor, device: torch.device
+) -> Tuple[torch.Tensor, torch.Tensor]:
+ """Packs DOM time series data into a pack sequence.
+
+ Also returns a tensor of x/y/z-coordinates and time of first/mean activation for each DOM.
+
+ Args:
+ tensor (torch.Tensor): Input tensor.
+ device (torch.device): Device to use.
+
+ Returns:
+ tensor (torch.Tensor): Packed sequence of DOM time series data.
+ xyztt (torch.Tensor): Tensor of x/y/z-coordinates as well as time of mean & first activations for each DOM.
+ """
+ tensor = sorted_jit_ignore(tensor)
+ xyztt = torch.stack(
+ [
+ torch.cat(
+ [
+ v[0, :3],
+ torch.as_tensor(
+ [v[:, 4].mean(), v[:, 4].min()], device=device
+ ),
+ ]
+ )
+ for v in tensor
+ ]
+ )
+
+ tensor = torch.nn.utils.rnn.pack_sequence(tensor, enforce_sorted=True)
+ return tensor, xyztt
+
+
+@torch.jit.ignore
+def sorted_jit_ignore(tensor: torch.Tensor) -> torch.Tensor:
+ """Sort a tensor based on the length of the elements.
+
+ Args:
+ tensor (torch.Tensor): Input tensor.
+
+ Returns:
+ torch.Tensor: Sorted tensor.
+ torch.Tensor: Indices of sorted tensor.
+ """
+ tensor = sorted(tensor, key=len, reverse=True)
+ return tensor
+
+
+@torch.jit.ignore
+def knn_graph_ignore(
+ x: torch.Tensor, k: int, batch: Optional[torch.Tensor] = None
+) -> torch.Tensor:
+ """Create a kNN graph based on the input data.
+
+ Args:
+ x: Input data.
+ k: Number of neighbours.
+ batch: Batch index.
+ """
+ return knn_graph(x=x, k=k, batch=batch)
diff --git a/src/graphnet/models/gnn/__init__.py b/src/graphnet/models/gnn/__init__.py
index 4da7f5b21..b5092e9fe 100644
--- a/src/graphnet/models/gnn/__init__.py
+++ b/src/graphnet/models/gnn/__init__.py
@@ -3,3 +3,4 @@
from .convnet import ConvNet
from .dynedge import DynEdge
from .dynedge_jinst import DynEdgeJINST
+from .node_rnn import Node_RNN
diff --git a/src/graphnet/models/gnn/node_rnn.py b/src/graphnet/models/gnn/node_rnn.py
new file mode 100644
index 000000000..1c5e1e969
--- /dev/null
+++ b/src/graphnet/models/gnn/node_rnn.py
@@ -0,0 +1,81 @@
+"""Implementation of the NodeTimeRNN model.
+
+(cannot be used as a standalone model)
+"""
+import torch
+
+from graphnet.models.gnn.gnn import GNN
+from graphnet.models.components.DOM_handling import (
+ append_dom_id,
+ DOM_time_series_to_pack_sequence,
+ DOM_to_time_series,
+ knn_graph_ignore,
+)
+from graphnet.utilities.config import save_model_config
+from torch_geometric.data import Data
+
+
+class Node_RNN(GNN):
+ """Implementation of the RNN model architecture.
+
+ The model takes as input the typical DOM data format and transforms it into
+ a time series of DOM activations pr. DOM. before applying a RNN layer and
+ outputting the an RNN output for each DOM.
+ """
+
+ @save_model_config
+ def __init__(
+ self,
+ nb_inputs: int,
+ hidden_size: int,
+ num_layers: int,
+ nb_neighbours: int,
+ ) -> None:
+ """Construct `NodeTimeRNN`.
+
+ Args:
+ nb_inputs: Number of features in the input data.
+ hidden_size: Number of features for the RNN output and hidden layers.
+ num_layers: Number of layers in the RNN.
+ nb_neighbours: Number of neighbours to use when reconstructing the graph representation.
+ """
+ self._nb_neighbours = nb_neighbours
+ self._num_layers = num_layers
+ self._hidden_size = hidden_size
+ self._nb_neighbours = nb_neighbours
+ self._nb_inputs = nb_inputs
+ super().__init__(nb_inputs, hidden_size + 5)
+
+ self._rnn = torch.nn.RNN(
+ num_layers=self._num_layers,
+ input_size=self._nb_inputs,
+ hidden_size=self._hidden_size,
+ )
+
+ def forward(self, data: Data) -> torch.Tensor:
+ """Apply learnable forward pass to the GNN."""
+ x, batch = data.x, data.batch
+
+ x = append_dom_id(x, batch, device=self.device) # add dom id to data
+ x, batch = DOM_to_time_series(
+ x, batch
+ ) # create time series for each dom
+ x, xyztt = DOM_time_series_to_pack_sequence(
+ x, device=self.device
+ ) # pack time series into pack sequence
+ x = self._rnn(x)[-1][0] # apply rnn layer
+
+ x = torch.hstack(
+ [xyztt, x]
+ ) # reintroduce x/y/z-coordinates and time of first/mean activation for each DOM
+
+ batch, sort_index = batch.sort()
+ data.x = x[sort_index]
+ data.batch = batch
+ edge_index = knn_graph_ignore(
+ x=data.x,
+ k=self._nb_neighbours,
+ batch=data.batch,
+ ).to(self.device)
+ data.edge_index = edge_index
+ return data
diff --git a/src/graphnet/models/gnn/rnn_dynedge.py b/src/graphnet/models/gnn/rnn_dynedge.py
new file mode 100644
index 000000000..e3e58b891
--- /dev/null
+++ b/src/graphnet/models/gnn/rnn_dynedge.py
@@ -0,0 +1,90 @@
+"""RNN_DynEdge model implementation."""
+from typing import List, Optional, Tuple, Union
+
+import torch
+from graphnet.models.gnn.gnn import GNN
+from graphnet.models.gnn.dynedge import DynEdge
+from graphnet.models.gnn.node_rnn import Node_RNN
+
+from graphnet.utilities.config import save_model_config
+from torch_geometric.data import Data
+
+
+class RNN_DynEdge(GNN):
+ """The RNN_DynEdge model class."""
+
+ @save_model_config
+ def __init__(
+ self,
+ nb_inputs: int,
+ *,
+ nb_neighbours: int = 8,
+ RNN_layers: int = 1,
+ RNN_hidden_size: int = 64,
+ features_subset: Optional[Union[List[int], slice]] = None,
+ dynedge_layer_sizes: Optional[List[Tuple[int, ...]]] = None,
+ post_processing_layer_sizes: Optional[List[int]] = None,
+ readout_layer_sizes: Optional[List[int]] = None,
+ global_pooling_schemes: Optional[Union[str, List[str]]] = None,
+ add_global_variables_after_pooling: bool = False,
+ ):
+ """Initialize the RNN_DynEdge model.
+
+ Args:
+ nb_inputs (int): Number of input features.
+ nb_neighbours (int, optional): Number of neighbours to consider.
+ Defaults to 8.
+ RNN_layers (int, optional): Number of RNN layers.
+ Defaults to 1.
+ RNN_hidden_size (int, optional): Size of the hidden state of the RNN. Also determines the size of the output of the RNN.
+ Defaults to 64.
+ features_subset (Optional[Union[List[int], slice]], optional): Subset of features to use.
+ dynedge_layer_sizes (Optional[List[Tuple[int, ...]]], optional): List of tuples of integers representing the sizes of the hidden layers of the DynEdge model.
+ post_processing_layer_sizes (Optional[List[int]], optional): List of integers representing the sizes of the hidden layers of the post-processing model.
+ readout_layer_sizes (Optional[List[int]], optional): List of integers representing the sizes of the hidden layers of the readout model.
+ global_pooling_schemes (Optional[Union[str, List[str]]], optional): Pooling schemes to use. Defaults to None.
+ add_global_variables_after_pooling (bool, optional): Whether to add global variables after pooling. Defaults to False.
+ """
+ self._nb_neighbours = nb_neighbours
+ self._nb_inputs = nb_inputs
+ self._RNN_layers = RNN_layers
+ self._RNN_hidden_size = RNN_hidden_size
+
+ self._feautres_subset = features_subset
+ self._dynedge_layer_sizes = dynedge_layer_sizes
+ self._post_processing_layer_sizes = post_processing_layer_sizes
+ self._global_pooling_schemes = global_pooling_schemes
+ self._add_global_variables_after_pooling = (
+ add_global_variables_after_pooling
+ )
+ if readout_layer_sizes is None:
+ readout_layer_sizes = [
+ 128,
+ ]
+ self._readout_layer_sizes = readout_layer_sizes
+
+ super().__init__(nb_inputs, self._readout_layer_sizes[-1])
+
+ self._rnn = Node_RNN(
+ num_layers=self._RNN_layers,
+ nb_inputs=self._nb_inputs,
+ hidden_size=self._RNN_hidden_size,
+ nb_neighbours=self._nb_neighbours,
+ )
+
+ self._dynedge = DynEdge(
+ nb_inputs=self._RNN_hidden_size + 5,
+ nb_neighbours=self._nb_neighbours,
+ dynedge_layer_sizes=self._dynedge_layer_sizes,
+ post_processing_layer_sizes=self._post_processing_layer_sizes,
+ readout_layer_sizes=self._readout_layer_sizes,
+ global_pooling_schemes=self._global_pooling_schemes,
+ add_global_variables_after_pooling=self._add_global_variables_after_pooling,
+ )
+
+ def forward(self, data: Data) -> torch.Tensor:
+ """Apply learnable forward pass of the RNN and DynEdge models."""
+ data = self._rnn(data)
+ readout = self._dynedge(data)
+
+ return readout
--
GitLab
From b88a0c005353cf8633d439672f511ab443283377 Mon Sep 17 00:00:00 2001
From: Aske-Rosted <askerosted@gmail.com>
Date: Mon, 12 Jun 2023 16:39:03 +0900
Subject: [PATCH 2/3] Change generation of Timeseries to dataset class
---
src/graphnet/data/dataset.py | 243 +++++++++++++++++-
src/graphnet/models/detector/icecube.py | 66 +++++
.../utilities/config/dataset_config.py | 1 +
3 files changed, 309 insertions(+), 1 deletion(-)
diff --git a/src/graphnet/data/dataset.py b/src/graphnet/data/dataset.py
index a0110988b..22a5c185c 100644
--- a/src/graphnet/data/dataset.py
+++ b/src/graphnet/data/dataset.py
@@ -17,6 +17,8 @@ from typing import (
import numpy as np
import torch
from torch_geometric.data import Data
+from torch.utils.data import TensorDataset
+
from graphnet.constants import GRAPHNET_ROOT_DIR
from graphnet.data.utilities.string_selection_resolver import (
@@ -150,6 +152,7 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
loss_weight_column: Optional[str] = None,
loss_weight_default_value: Optional[float] = None,
seed: Optional[int] = None,
+ timeseries: Optional[bool] = None,
):
"""Construct Dataset.
@@ -195,6 +198,7 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
subset of events when resolving a string-based selection (e.g.,
`"10000 random events ~ event_no % 5 > 0"` or `"20% random
events ~ event_no % 5 > 0"`).
+ timeseries: Whether the dataset is a timeseries dataset. Defaults to None (Not a timeseries dataset).
"""
# Base class constructor
super().__init__(name=__name__, class_name=self.__class__.__name__)
@@ -218,6 +222,7 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
self._index_column = index_column
self._truth_table = truth_table
self._loss_weight_default_value = loss_weight_default_value
+ self._timeseries = timeseries
if node_truth is not None:
assert isinstance(node_truth_table, str)
@@ -375,7 +380,14 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
features, truth, node_truth, loss_weight = self._query(
sequential_index
)
- graph = self._create_graph(features, truth, node_truth, loss_weight)
+ if self._timeseries is not False:
+ graph = self._create_timeseries(
+ features, truth, node_truth, loss_weight
+ )
+ else:
+ graph = self._create_graph(
+ features, truth, node_truth, loss_weight
+ )
return graph
# Internal method(s)
@@ -616,6 +628,223 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
graph["dataset_path"] = self._path
return graph
+ def _create_timeseries(
+ self,
+ features: List[Tuple[float, ...]],
+ truth: Tuple[Any, ...],
+ node_truth: Optional[List[Tuple[Any, ...]]] = None,
+ loss_weight: Optional[float] = None,
+ ) -> Data:
+ """Create Pytorch Data (i.e. timeseries) object.
+
+ No preprocessing is performed at this stage, just as no node adjancency
+ is imposed. This means that the `edge_attr` and `edge_weight`
+ attributes are not set.
+
+ Args:
+ features: List of list of tuples, containing event features.
+ truth: List of tuples, containing truth information.
+ node_truth: List of tuples, containing node-level truth.
+ loss_weight: A weight associated with the event for weighing the
+ loss.
+
+ Returns:
+ time series object.
+ """
+ # Convert nested list to simple dict
+ truth_dict = {
+ key: truth[index] for index, key in enumerate(self._truth)
+ }
+
+ # Define custom labels
+ labels_dict = self._get_labels(truth_dict)
+
+ # Convert nested list to simple dict
+ if node_truth is not None:
+ node_truth_array = np.asarray(node_truth)
+ assert self._node_truth is not None
+ node_truth_dict = {
+ key: node_truth_array[:, index]
+ for index, key in enumerate(self._node_truth)
+ }
+
+ # Catch cases with no reconstructed pulses
+ if len(features):
+ data = np.asarray(features)[:, 1:]
+ else:
+ data = np.array([]).reshape((0, len(self._features) - 1))
+
+ # Construct graph data object
+ x = torch.tensor(data, dtype=self._dtype) # pylint: disable=C0103
+ n_pulses = torch.tensor(len(x), dtype=torch.long)
+ if "dom_number" not in features:
+ number_index = -1
+ unique_index = [
+ self._features.index(key) - 1
+ for key in ["dom_x", "dom_y", "dom_z"]
+ ]
+ dom_number = torch.unique(
+ x[:, unique_index], return_inverse=True, dim=0, sorted=False
+ )[-1]
+ x = torch.cat([x, dom_number.reshape(-1, 1)], dim=1)
+ else:
+ number_index = self._features.index("dom_number") - 1
+
+ dom_activation_sort = x[:, number_index].sort()[-1]
+ x = x[dom_activation_sort]
+ bin_count = torch.bincount(
+ x[:, number_index].type(torch.int64)
+ ).cumsum(0)
+ # removing dom_number from features should maybe be optional.
+ if number_index != -1:
+ x = x[:, torch.arange(x.shape[-1]) != number_index]
+ else:
+ x = x[:, :-1]
+ n_activations = bin_count - torch.cat(
+ [torch.tensor([0]), bin_count[:-1]]
+ )
+ activations_sort = n_activations.sort(descending=True)[-1]
+ x = np.array(
+ [np.array(_) for _ in torch.tensor_split(x, bin_count[:-1])],
+ dtype=object,
+ )[activations_sort]
+ x = [
+ torch.tensor(
+ dom_list[
+ np.argsort(
+ dom_list[
+ :,
+ self._features.index("dom_time"),
+ ]
+ )[::-1]
+ ]
+ )
+ for dom_list in x
+ ]
+ x, xyztt = Dataset.DOM_time_series_to_pack_sequence(
+ x, device="cpu", features=self._features[1:]
+ )
+ graph = TimeSeries(x=xyztt, edge_index=None)
+ graph.time_series = [
+ x.data,
+ x.batch_sizes,
+ x.sorted_indices,
+ x.unsorted_indices,
+ ]
+ # print(x.batch_sizes)
+ graph.n_pulses = n_pulses
+ graph.features = self._features[1:]
+ # graph.n_activations = n_activations[activations_sort]
+
+ # Add loss weight to graph.
+ if loss_weight is not None and self._loss_weight_column is not None:
+ # No loss weight was retrieved, i.e., it is missing for the current
+ # event.
+ if loss_weight < 0:
+ if self._loss_weight_default_value is None:
+ raise ValueError(
+ "At least one event is missing an entry in "
+ f"{self._loss_weight_column} "
+ "but loss_weight_default_value is None."
+ )
+ graph[self._loss_weight_column] = torch.tensor(
+ self._loss_weight_default_value, dtype=self._dtype
+ ).reshape(-1, 1)
+ else:
+ graph[self._loss_weight_column] = torch.tensor(
+ loss_weight, dtype=self._dtype
+ ).reshape(-1, 1)
+
+ # Write attributes, either target labels, truth info or original
+ # features.
+ add_these_to_graph = [labels_dict, truth_dict]
+ if node_truth is not None:
+ add_these_to_graph.append(node_truth_dict)
+ for write_dict in add_these_to_graph:
+ for key, value in write_dict.items():
+ try:
+ graph[key] = torch.tensor(value)
+ except TypeError:
+ # Cannot convert `value` to Tensor due to its data type,
+ # e.g. `str`.
+ self.debug(
+ (
+ f"Could not assign `{key}` with type "
+ f"'{type(value).__name__}' as attribute to graph."
+ )
+ )
+
+ # Additionally add original features as (static) attributes
+ for index, feature in enumerate(graph.features[:-2]):
+ if feature not in ["x"]:
+ graph[feature] = graph.x[:, index]
+
+ # Add custom labels to the graph
+ for key, fn in self._label_fns.items():
+ graph[key] = fn(graph)
+
+ # Add Dataset Path. Useful if multiple datasets are concatenated.
+ graph["dataset_path"] = self._path
+ return graph
+
+ def DOM_time_series_to_pack_sequence(
+ tensor: torch.Tensor, device: torch.device, features: List[str]
+ ) -> Tuple[torch.Tensor, torch.Tensor]:
+ """Packs DOM time series data into a pack sequence.
+
+ Also returns a tensor of x/y/z-coordinates and time of first/mean activation for each DOM.
+
+ Args:
+ tensor (torch.Tensor): Input tensor.
+ device (torch.device): Device to use.
+ features (list): List of features in the tensor.
+
+ Returns:
+ tensor (torch.Tensor): Packed sequence of DOM time series data.
+ xyztt (torch.Tensor): Tensor of x/y/z-coordinates as well as time of mean & first activations for each DOM.
+ """
+ xyztc = torch.stack(
+ [
+ torch.cat(
+ [
+ v[
+ 0,
+ [
+ features.index("dom_x"),
+ features.index("dom_y"),
+ features.index("dom_z"),
+ ],
+ ],
+ torch.as_tensor(
+ [
+ v[:, features.index("dom_time")].mean(),
+ v[:, features.index("charge")].sum(),
+ ],
+ device=device,
+ ),
+ ]
+ )
+ for v in tensor
+ ]
+ )
+
+ tensor = torch.nn.utils.rnn.pack_sequence(tensor, enforce_sorted=True)
+ return tensor, xyztc
+
+ @torch.jit.ignore
+ def sorted_jit_ignore(tensor: torch.Tensor) -> torch.Tensor:
+ """Sort a tensor based on the length of the elements.
+
+ Args:
+ tensor (torch.Tensor): Input tensor.
+
+ Returns:
+ torch.Tensor: Sorted tensor.
+ torch.Tensor: Indices of sorted tensor.
+ """
+ tensor = sorted(tensor, key=len, reverse=True)
+ return tensor
+
def _get_labels(self, truth_dict: Dict[str, Any]) -> Dict[str, Any]:
"""Return dictionary of labels, to be added as graph attributes."""
if "pid" in truth_dict.keys():
@@ -664,6 +893,18 @@ class Dataset(Logger, Configurable, torch.utils.data.Dataset, ABC):
return -1
+class TimeSeries(Data):
+ """Dataset class for time series data."""
+
+ def __cat_dim__(
+ self: Data, key: str, value: Any, *args: Any, **kwargs: Dict[str, Any]
+ ) -> Any:
+ """Specify change in concatenation dimension for specific key."""
+ if key == "time_series":
+ return 0
+ return super().__cat_dim__(key, value, *args, **kwargs)
+
+
class EnsembleDataset(torch.utils.data.ConcatDataset):
"""Construct a single dataset from a collection of datasets."""
diff --git a/src/graphnet/models/detector/icecube.py b/src/graphnet/models/detector/icecube.py
index e6addb8e3..4bd162042 100644
--- a/src/graphnet/models/detector/icecube.py
+++ b/src/graphnet/models/detector/icecube.py
@@ -71,6 +71,72 @@ class IceCubeKaggle(Detector):
return data
+class IceCube86RNN(Detector):
+ """`Detector` class for IceCube-86 with a time-series data input."""
+
+ features = FEATURES.ICECUBE86[:5]
+
+ def _forward(self, data: Data) -> Data:
+ """Ingest data, build graph, and preprocess features.
+
+ Args:
+ data: Input graph data.
+
+ Returns:
+ Connected and preprocessed time-series graph data.
+ """
+ self._validate_features(data)
+
+ data.x[:, 0] /= 500.0 # dom_x
+ data.x[:, 1] /= 500.0 # dom_y
+ data.x[:, 2] /= 500.0 # dom_z
+ data.x[:, 3] = (data.x[:, 3] - 1.0e04) / 3.0e4 # dom_time
+ data.x[:, 4] = torch.log10(data.x[:, 4]) / 3.0 # charge
+
+ data.time_series[0] /= torch.tensor(
+ [500.0, 500.0, 500.0, 1.0, 1.0], device=data.x.device
+ )
+ data.time_series[0][:, 3] = (
+ data.time_series[0][:, 3] - 1.0e04
+ ) / 3.0e4
+ data.time_series[0][:, 4] = (
+ torch.log10(data.time_series[0][:, 4]) / 3.0
+ )
+
+ data.time_series[0] = torch.tensor_split(
+ data.time_series[0], (data.n_pulses.cumsum(0)[:-1]).cpu()
+ )
+
+ data.time_series[1] = torch.tensor_split(
+ data.time_series[1],
+ (
+ torch.argwhere(
+ torch.gt(data.time_series[1][1:], data.time_series[1][:-1])
+ ).flatten()
+ + 1
+ ).cpu(),
+ )
+
+ time_series = []
+ for sequence, batch_sizes, sorted_indices, unsorted_indices in zip(
+ data.time_series[0],
+ data.time_series[1],
+ data.time_series[2],
+ data.time_series[3],
+ ):
+ time_series.append(
+ torch.nn.utils.rnn.PackedSequence(
+ data=sequence,
+ batch_sizes=batch_sizes.cpu(),
+ sorted_indices=sorted_indices,
+ unsorted_indices=unsorted_indices,
+ )
+ )
+
+ data.time_series = time_series
+ return data
+
+
class IceCubeDeepCore(IceCube86):
"""`Detector` class for IceCube-DeepCore."""
diff --git a/src/graphnet/utilities/config/dataset_config.py b/src/graphnet/utilities/config/dataset_config.py
index fc5beb44b..782ff7c6f 100644
--- a/src/graphnet/utilities/config/dataset_config.py
+++ b/src/graphnet/utilities/config/dataset_config.py
@@ -47,6 +47,7 @@ class DatasetConfig(BaseConfig):
loss_weight_default_value: Optional[float] = None
seed: Optional[int] = None
+ timeseries: Optional[bool] = None
def __init__(self, **data: Any) -> None:
"""Construct `DataConfig`.
--
GitLab
From 5a03cb10f3d203d60aee5cba8c9b1c507dd6546f Mon Sep 17 00:00:00 2001
From: Aske-Rosted <askerosted@gmail.com>
Date: Mon, 12 Jun 2023 16:41:52 +0900
Subject: [PATCH 3/3] Update RNN to new dataset
---
src/graphnet/models/gnn/node_rnn.py | 31 ++++++++------------------
src/graphnet/models/gnn/rnn_dynedge.py | 2 +-
2 files changed, 10 insertions(+), 23 deletions(-)
diff --git a/src/graphnet/models/gnn/node_rnn.py b/src/graphnet/models/gnn/node_rnn.py
index 1c5e1e969..490c1f024 100644
--- a/src/graphnet/models/gnn/node_rnn.py
+++ b/src/graphnet/models/gnn/node_rnn.py
@@ -54,28 +54,15 @@ class Node_RNN(GNN):
def forward(self, data: Data) -> torch.Tensor:
"""Apply learnable forward pass to the GNN."""
- x, batch = data.x, data.batch
+ rnn_out = []
+ for b_uniq in data.batch.unique():
+ rnn_out.append(
+ self._rnn(data.time_series[b_uniq])[-1][0]
+ ) # apply rnn layer
+ # x = self._rnn(x)[-1][0] # apply rnn layer
+ rnn_out = torch.cat(rnn_out)
- x = append_dom_id(x, batch, device=self.device) # add dom id to data
- x, batch = DOM_to_time_series(
- x, batch
- ) # create time series for each dom
- x, xyztt = DOM_time_series_to_pack_sequence(
- x, device=self.device
- ) # pack time series into pack sequence
- x = self._rnn(x)[-1][0] # apply rnn layer
-
- x = torch.hstack(
- [xyztt, x]
+ data.x = torch.hstack(
+ [data.x, rnn_out]
) # reintroduce x/y/z-coordinates and time of first/mean activation for each DOM
-
- batch, sort_index = batch.sort()
- data.x = x[sort_index]
- data.batch = batch
- edge_index = knn_graph_ignore(
- x=data.x,
- k=self._nb_neighbours,
- batch=data.batch,
- ).to(self.device)
- data.edge_index = edge_index
return data
diff --git a/src/graphnet/models/gnn/rnn_dynedge.py b/src/graphnet/models/gnn/rnn_dynedge.py
index e3e58b891..0deec7113 100644
--- a/src/graphnet/models/gnn/rnn_dynedge.py
+++ b/src/graphnet/models/gnn/rnn_dynedge.py
@@ -59,7 +59,7 @@ class RNN_DynEdge(GNN):
)
if readout_layer_sizes is None:
readout_layer_sizes = [
- 128,
+ 512,
]
self._readout_layer_sizes = readout_layer_sizes
--
GitLab