wip neural arquitechture search

flake.nix

@@ -26,6 +26,7 @@
           #mujoco = callPackage ./nix/mujoco.nix { };
           trimesh = callPackage ./nix/trimesh.nix { };
           brax = callPackage ./nix/brax.nix { };
+          mplcursors = callPackage ./nix/mplcursors.nix { };
         };
 
         core-python = pkgs.python3.withPackages (ps: with ps; [
@@ -34,15 +35,18 @@
           pandas
           matplotlib
           pytorch
+          torchvision
           pytest
           tqdm
           rich
+          networkx
           # wandb # test failing
 
           # only supported on linux
           jaxlib
           jax
           overlay.brax
+          overlay.mplcursors
         ]);
 
 

nix/mplcursors.nix

@@ -0,0 +1,31 @@
+{ lib
+, buildPythonPackage
+, fetchPypi
+, setuptools_scm
+, pytest
+, pytestCheckHook
+, matplotlib
+}:
+
+buildPythonPackage rec {
+  pname = "mplcursors";
+  version = "0.5";
+
+  src = fetchPypi {
+    inherit pname version;
+    hash = "sha256-w92Ej+z0b7xIdAikBgouEDpVZ8EERN8OrIswjEBU5U0=";
+  };
+
+  nativeBuildInputs = [ setuptools_scm ];
+  propagatedBuildInputs = [ matplotlib ];
+  checkInputs = [ pytest pytestCheckHook ];
+
+  doCheck = false;
+
+  meta = with lib; {
+    description = "Interactive data selection cursors for Matplotlib";
+    homepage = "https://github.com/anntzer/mplcursors";
+    license = licenses.mit;
+    maintainers = with maintainers; [ ];
+  };
+}

src/mnist/main.py

@@ -0,0 +1,201 @@
+from __future__ import print_function
+
+import argparse
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.optim as optim
+from torch.optim.lr_scheduler import StepLR
+from torchvision import datasets, transforms
+
+
+class Net(nn.Module):
+    def __init__(self):
+        super(Net, self).__init__()
+        self.conv1 = nn.Conv2d(1, 32, 3, 1)
+        self.conv2 = nn.Conv2d(32, 64, 3, 1)
+        self.dropout1 = nn.Dropout(0.25)
+        self.dropout2 = nn.Dropout(0.5)
+        self.fc1 = nn.Linear(9216, 128)
+        self.fc2 = nn.Linear(128, 10)
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = F.relu(x)
+        x = self.conv2(x)
+        x = F.relu(x)
+        x = F.max_pool2d(x, 2)
+        x = self.dropout1(x)
+        x = torch.flatten(x, 1)
+        x = self.fc1(x)
+        x = F.relu(x)
+        x = self.dropout2(x)
+        x = self.fc2(x)
+        output = F.log_softmax(x, dim=1)
+        return output
+
+
+def train(args, model, device, train_loader, optimizer, epoch):
+    model.train()
+    for batch_idx, (data, target) in enumerate(train_loader):
+        data, target = data.to(device), target.to(device)
+        optimizer.zero_grad()
+        output = model(data)
+        loss = F.nll_loss(output, target)
+        loss.backward()
+        optimizer.step()
+        if batch_idx % args.log_interval == 0:
+            print(
+                "Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}".format(
+                    epoch,
+                    batch_idx * len(data),
+                    len(train_loader.dataset),
+                    100.0 * batch_idx / len(train_loader),
+                    loss.item(),
+                )
+            )
+            if args.dry_run:
+                break
+
+
+def test(model, device, test_loader):
+    model.eval()
+    test_loss = 0
+    correct = 0
+    with torch.no_grad():
+        for data, target in test_loader:
+            data, target = data.to(device), target.to(device)
+            output = model(data)
+            test_loss += F.nll_loss(
+                output, target, reduction="sum"
+            ).item()  # sum up batch loss
+            pred = output.argmax(
+                dim=1, keepdim=True
+            )  # get the index of the max log-probability
+            correct += pred.eq(target.view_as(pred)).sum().item()
+
+    test_loss /= len(test_loader.dataset)
+
+    print(
+        "\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n".format(
+            test_loss,
+            correct,
+            len(test_loader.dataset),
+            100.0 * correct / len(test_loader.dataset),
+        )
+    )
+
+
+def main():
+    # Training settings
+    parser = argparse.ArgumentParser(description="PyTorch MNIST Example")
+    parser.add_argument(
+        "--batch-size",
+        type=int,
+        default=64,
+        metavar="N",
+        help="input batch size for training (default: 64)",
+    )
+    parser.add_argument(
+        "--test-batch-size",
+        type=int,
+        default=1000,
+        metavar="N",
+        help="input batch size for testing (default: 1000)",
+    )
+    parser.add_argument(
+        "--epochs",
+        type=int,
+        default=14,
+        metavar="N",
+        help="number of epochs to train (default: 14)",
+    )
+    parser.add_argument(
+        "--lr",
+        type=float,
+        default=1.0,
+        metavar="LR",
+        help="learning rate (default: 1.0)",
+    )
+    parser.add_argument(
+        "--gamma",
+        type=float,
+        default=0.7,
+        metavar="M",
+        help="Learning rate step gamma (default: 0.7)",
+    )
+    parser.add_argument(
+        "--no-cuda", action="store_true", default=False, help="disables CUDA training"
+    )
+    parser.add_argument(
+        "--no-mps",
+        action="store_true",
+        default=False,
+        help="disables macOS GPU training",
+    )
+    parser.add_argument(
+        "--dry-run",
+        action="store_true",
+        default=False,
+        help="quickly check a single pass",
+    )
+    parser.add_argument(
+        "--seed", type=int, default=1, metavar="S", help="random seed (default: 1)"
+    )
+    parser.add_argument(
+        "--log-interval",
+        type=int,
+        default=10,
+        metavar="N",
+        help="how many batches to wait before logging training status",
+    )
+    parser.add_argument(
+        "--save-model",
+        action="store_true",
+        default=False,
+        help="For Saving the current Model",
+    )
+    args = parser.parse_args()
+    use_cuda = not args.no_cuda and torch.cuda.is_available()
+    use_mps = not args.no_mps and torch.backends.mps.is_available()
+
+    torch.manual_seed(args.seed)
+
+    if use_cuda:
+        device = torch.device("cuda")
+    elif use_mps:
+        device = torch.device("mps")
+    else:
+        device = torch.device("cpu")
+
+    train_kwargs = {"batch_size": args.batch_size}
+    test_kwargs = {"batch_size": args.test_batch_size}
+    if use_cuda:
+        cuda_kwargs = {"num_workers": 1, "pin_memory": True, "shuffle": True}
+        train_kwargs.update(cuda_kwargs)
+        test_kwargs.update(cuda_kwargs)
+
+    transform = transforms.Compose(
+        [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]
+    )
+    dataset1 = datasets.MNIST("../data", train=True, download=True, transform=transform)
+    dataset2 = datasets.MNIST("../data", train=False, transform=transform)
+    train_loader = torch.utils.data.DataLoader(dataset1, **train_kwargs)
+    test_loader = torch.utils.data.DataLoader(dataset2, **test_kwargs)
+
+    model = Net().to(device)
+    optimizer = optim.Adadelta(model.parameters(), lr=args.lr)
+
+    scheduler = StepLR(optimizer, step_size=1, gamma=args.gamma)
+    for epoch in range(1, args.epochs + 1):
+        train(args, model, device, train_loader, optimizer, epoch)
+        test(model, device, test_loader)
+        scheduler.step()
+
+    if args.save_model:
+        torch.save(model.state_dict(), "mnist_cnn.pt")
+
+
+if __name__ == "__main__":
+    main()

src/module.py

@@ -0,0 +1,57 @@
+class Region:
+    """
+    This is the basic unit of computation which evolves through:
+    - Randon mutation
+    - A global reward signal
+    - Backpropagation
+    """
+
+    def __init__(self) -> None:
+
+        self._input = None
+        # the module optionally receives a tensor as an input. The module will
+        # define  it's architecture to be able to process any input shape.
+
+        self._reward = None
+        # the module learnable parameters can be updated through
+        # backpropagation and from a global reward signal.
+        # this must be a scalar
+
+        self._gradient = None
+        # The gradient used to update the parameters
+        # this must have the same shape as output
+
+        self._mutation = 0.5
+        # [0,1] this is the global mutation rate which affects how quickly the
+        # module evolves. zero means that the module is frozen, one means that
+        # the module changes in every opportunity.
+
+    def __str__(self):
+        return f"""
+        Region:
+            Mutation rate: {self._mutation}
+        """
+
+
+class Brain:
+    """
+    This is a directed acyclic graph of `MutableModule`.
+    It randomly creates new nodes and edges between them.
+    It also responsible of making the shapes work out.
+
+    We'll probably want a Tensorboard-like visualization of this graph.
+    """
+
+    def __init__(self) -> None:
+
+        self._mutation = 0.5
+        # [0,1] this is the global mutation rate which affects how quickly the
+        # module evolves. zero means that the module is frozen, one means that
+        # the module changes in every opportunity.
+
+
+# The simplest thing we can implement is a fully connected network were we mutate the weights
+# and that's the only learning method
+
+if __name__ == "__main__":
+    print(Region())

src/nas/evolve.py

@@ -0,0 +1,32 @@
+import random
+from copy import deepcopy
+from typing import Callable
+
+from nas import Net
+
+
+def select_parents(population, fitness_scores):
+    """
+    Select the top 20% fittest individuals in the population
+    """
+    sorted_population = sorted(
+        population,
+        key=lambda individual: fitness_scores[population.index(individual)],
+        reverse=False,
+    )
+    return sorted_population[: len(sorted_population) // 5]
+
+
+def evolve(constructor, fitness_fn: Callable, population_size: int, steps: int):
+    """ """
+    population = [constructor() for _ in range(population_size)]
+    for i in range(steps):
+        fitness_scores = [fitness_fn(individual) for individual in population]
+        print(fitness_scores)
+        next_gen = select_parents(population, fitness_scores)
+        while len(next_gen) < population_size:
+            individual = deepcopy(random.choice(next_gen))
+            next_gen.append(individual.mutate())
+        population = next_gen
+
+    return min(population, key=fitness_fn)