Merge pull request #4 from dirac-institute/awo/initial-model

Adding example CNN to the set of models

example_config.toml

@@ -0,0 +1,72 @@
+[general]
+use_gpu = true
+
+# Destination of log messages
+# 'stderr' and 'stdout' specify the console.
+log_destination = "stderr"
+# A path name specifies a file e.g.
+# log = "fibad_log.txt"
+
+# Lowest log level to emit.
+# As you go down the list, fibad will become more verbose in the log.
+#
+# log_level = "critical" # Only emit the most severe of errors
+# log_level = "error"    # Emit all errors
+# log_level = "warning"  # Emit warnings and all errors
+log_level = "info"     # Emit informational messages, warnings and all errors
+# log_level = "debug"    # Very verbose, emit all log messages.
+
+[download]
+sw = "22asec"
+sh = "22asec"
+filter = ["HSC-G", "HSC-R", "HSC-I", "HSC-Z", "HSC-Y"]
+type = "coadd"
+rerun = "pdr3_wide"
+username = "mtauraso@local"
+password = "cCw+nX53lmNLHMy+JbizpH/dl4t7sxljiNm6a7k1"
+max_connections = 2
+fits_file = "../hscplay/temp.fits"
+cutout_dir = "../hscplay/cutouts/"
+offset = 0
+num_sources = 500
+
+# These control the downloader's HTTP requests and retries
+# `retry_wait` How long to wait before retrying a failed HTTP request in seconds. Default 30s
+retry_wait = 30
+# `retries` How many times to retry a failed HTTP request before moving on to the next one. Default 3 times
+retries = 3
+# `timepout` How long should we wait to get a full HTTP response from the server. Default 3600s (1hr)
+timeout = 3600
+# `chunksize` How many sky location rectangles should we request in a single request. Default is 990
+chunksize = 990
+
+[model]
+# name = "ExampleCNN"
+# name = "ExampleAutoencoder"
+
+# An example of requesting an external model class
+# external_class = "user_package.submodule.ExternalModel"
+external_cls = "kbmod_ml.models.cnn.CNN"
+
+weights_filepath = "example_model.pth"
+epochs = 10
+
+[data_loader]
+# Name of data loader to use
+name = "CifarDataLoader"
+# name = "HSCDataLoader"
+
+# An example of requesting an external data loader class
+# external_class = "user_package.submodule.ExternalDataLoader"
+
+# Directory path where the data is stored
+path = "/Users/drew/code/fibad/data/cifar-10-batches-py"
+# path = "/Users/drew/code/fibad/data/hsc-samples"
+
+# Default PyTorch DataLoader parameters
+batch_size = 10
+shuffle = true
+num_workers = 10
+
+[predict]
+batch_size = 32

pyproject.toml

@@ -16,6 +16,8 @@ classifiers = [
 dynamic = ["version"]
 requires-python = ">=3.9"
 dependencies = [
+    "torch", # PyTorch
+    # "fibad", when it is available on PyPI
 ]
 
 [project.urls]

src/kbmod_ml/models/cnn.py

@@ -0,0 +1,74 @@
+# ruff: noqa: D101, D102
+
+# This example model is taken from the PyTorch CIFAR10 tutorial:
+# https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html#define-a-convolutional-neural-network
+import logging
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F  # noqa N812
+import torch.optim as optim
+from fibad.models.model_registry import fibad_model
+
+logger = logging.getLogger(__name__)
+
+
+@fibad_model
+class CNN(nn.Module):
+    def __init__(self, model_config, shape):
+        logger.info("This is an external model, not in FIBAD!!!")
+        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.config = model_config
+
+        # Optimizer and criterion could be set directly, i.e. `self.optimizer = optim.SGD(...)`
+        # but we define them as methods as a way to allow for more flexibility in the future.
+        self.optimizer = self._optimizer()
+        self.criterion = self._criterion()
+
+    def forward(self, x):
+        x = self.pool(F.relu(self.conv1(x)))
+        x = self.pool(F.relu(self.conv2(x)))
+        x = torch.flatten(x, 1)
+        x = F.relu(self.fc1(x))
+        x = F.relu(self.fc2(x))
+        x = self.fc3(x)
+        return x
+
+    def train_step(self, batch):
+        """This function contains the logic for a single training step. i.e. the
+        contents of the inner loop of a ML training process.
+
+        Parameters
+        ----------
+        batch : tuple
+            A tuple containing the inputs and labels for the current batch.
+
+        Returns
+        -------
+        Current loss value
+            The loss value for the current batch.
+        """
+        inputs, labels = batch
+
+        self.optimizer.zero_grad()
+        outputs = self(inputs)
+        loss = self.criterion(outputs, labels)
+        loss.backward()
+        self.optimizer.step()
+        return {"loss": loss.item()}
+
+    def _criterion(self):
+        return nn.CrossEntropyLoss()
+
+    def _optimizer(self):
+        return optim.SGD(self.parameters(), lr=0.001, momentum=0.9)
+
+    def save(self):
+        torch.save(self.state_dict(), self.config.get("weights_filepath", "example_cnn.pth"))