Merge pull request #149 from geometric-intelligence/analyze_dual_agents

Run topo vae, Analyze dual agents

.gitignore

@@ -1,5 +1,11 @@
 # API KEY
-neurometry/api_key.txt
+*api_key.txt
+
+neurometry/results/*
+neurometry/wandb/*
+
+neurometry/datasets/rnn_grid_cells/Dual agent path integration high res/*
+neurometry/datasets/rnn_grid_cells/Single agent path integration high res/*
 
 
 *viewer*

neurometry/curvature/default_config.py

@@ -11,7 +11,8 @@
 # WANDB API KEY
 # Find it here: https://wandb.ai/authorize
 # Story it in file: api_key.txt (without extra line break)
-with open("api_key.txt") as f:
+api_key_path = os.path.join(os.getcwd(), "curvature","api_key.txt")
+with open(api_key_path) as f:
     api_key = f.read()
 
 # Directories
@@ -139,7 +140,7 @@
 
 # Datasets
 # dataset_name = ["s1_synthetic", "s2_synthetic"]
-dataset_name = ["kb_synthetic"]
+dataset_name = ["s1_synthetic"]
 for one_dataset_name in dataset_name:
     if one_dataset_name not in [
         "s1_synthetic",
@@ -163,10 +164,10 @@
 
 # Only used of dataset_name in ["s1_synthetic", "s2_synthetic", "t2_synthetic"]
 n_times = [2500]  # , 2000]  # actual number of times is sqrt_ntimes ** 2
-embedding_dim = [3, 10, 20, 30]  # for s1 stopped at 5 (not done, but 3 was done)
+embedding_dim = [5]  # for s1 stopped at 5 (not done, but 3 was done)
 geodesic_distortion_amp = [0.4]
 # TODO: Add 0.03, possibly 0,000[1
-noise_var = [0.1, 0.075, 0.05, 0.03, 0.01, 0.005, 0.001]  # , 1e-2, 1e-1] 0.075, 0.1] #[
+noise_var = [0.1]  # , 1e-2, 1e-1] 0.075, 0.1] #[
 
 # Only used if dataset_name == "grid_cells"
 grid_scale = [1.0]

neurometry/curvature/evaluate.py

@@ -8,7 +8,7 @@
 import geomstats.backend as gs  # noqa: E402
 
 # import gph
-from datasets.synthetic import (  # noqa: E402
+from neurometry.curvature.datasets.synthetic import (  # noqa: E402
     get_s1_synthetic_immersion,
     get_s2_synthetic_immersion,
     get_t2_synthetic_immersion,
@@ -106,7 +106,7 @@ def get_z_grid(config, n_grid_points=100):
         z_grid = torch.cartesian_prod(thetas, phis)
     return z_grid
 
-
+#TODO: change instantiation of PullbackMetric to match latest geomstats version
 def _compute_curvature(z_grid, immersion, dim, embedding_dim):
     """Compute mean curvature vector and its norm at each point."""
     neural_metric = PullbackMetric(

neurometry/curvature/losses.py

@@ -67,6 +67,8 @@ def elbo(x, x_mu, posterior_params, z, labels, config):
 
     if config.gen_likelihood_type == "gaussian":
         recon_loss = torch.mean((x - x_mu).pow(2))
+    else:
+        raise NotImplementedError
 
     if config.dataset_name == "s1_synthetic":
         recon_loss = recon_loss / (config.radius**2)

neurometry/curvature/main.py

@@ -14,21 +14,21 @@
 
 # from ray.tune.integration.wandb import wandb_mixin
 import torch
-import train
-import viz
+import neurometry.curvature.train as train
+import neurometry.curvature.viz as viz
 import wandb
 from ray import air, tune
 from ray.tune.schedulers import AsyncHyperBandScheduler
 from ray.tune.search.hyperopt import HyperOptSearch
 
 os.environ["GEOMSTATS_BACKEND"] = "pytorch"
-import datasets.utils  # noqa: E402
-import default_config  # noqa: E402
-import evaluate  # noqa: E402
+import neurometry.curvature.datasets.utils as utils  # noqa: E402
+import neurometry.curvature.default_config as default_config # noqa: E402
+import neurometry.curvature.evaluate as evaluate # noqa: E402
 import geomstats.backend as gs  # noqa: E402
-import models.klein_bottle_vae  # noqa: E402
-import models.neural_vae  # noqa: E402
-import models.toroidal_vae  # noqa: E402
+import neurometry.curvature.models.klein_bottle_vae as klein_bottle_vae  # noqa: E402
+import neurometry.curvature.models.neural_vae as neural_vae # noqa: E402
+import neurometry.curvature.models.toroidal_vae as toroidal_vae # noqa: E402
 
 # Required to make matplotlib figures in threads:
 matplotlib.use("Agg")
@@ -262,7 +262,7 @@ def main_run(sweep_config):
         wandb.run.name = run_name
 
         # Load data, labels
-        dataset, labels, train_loader, test_loader = datasets.utils.load(wandb_config)
+        dataset, labels, train_loader, test_loader = utils.load(wandb_config)
         data_n_times, data_dim = dataset.shape
         wandb_config.update(
             {
@@ -344,7 +344,7 @@ def create_model_and_train_test(config, train_loader, test_loader):
         torch.manual_seed(0)
         if torch.cuda.is_available():
             torch.cuda.manual_seed_all(0)
-        model = models.neural_vae.NeuralVAE(
+        model = neural_vae.NeuralVAE(
             data_dim=data_dim,
             latent_dim=config.latent_dim,
             sftbeta=config.sftbeta,
@@ -360,7 +360,7 @@ def create_model_and_train_test(config, train_loader, test_loader):
         torch.manual_seed(0)
         if torch.cuda.is_available():
             torch.cuda.manual_seed_all(0)
-        model = models.toroidal_vae.ToroidalVAE(
+        model = toroidal_vae.ToroidalVAE(
             data_dim=data_dim,
             latent_dim=config.latent_dim,
             sftbeta=config.sftbeta,
@@ -375,7 +375,7 @@ def create_model_and_train_test(config, train_loader, test_loader):
         torch.manual_seed(0)
         if torch.cuda.is_available():
             torch.cuda.manual_seed_all(0)
-        model = models.klein_bottle_vae.KleinBottleVAE(
+        model = klein_bottle_vae.KleinBottleVAE(
             data_dim=data_dim,
             latent_dim=config.latent_dim,
             sftbeta=config.sftbeta,
@@ -587,4 +587,4 @@ def curvature_compute_plot_log(config, dataset, labels, model):
     plt.close("all")
 
 
-main()
+#main()

neurometry/curvature/models/klein_bottle_vae.py

@@ -5,7 +5,7 @@
 
 import geomstats.backend as gs
 import torch
-from hyperspherical.distributions import VonMisesFisher
+from neurometry.curvature.hyperspherical.distributions.von_mises_fisher import VonMisesFisher
 from torch.nn import functional as F
 
 

neurometry/curvature/models/neural_vae.py

@@ -4,7 +4,7 @@
 """
 
 import torch
-from hyperspherical.distributions import VonMisesFisher
+from neurometry.curvature.hyperspherical.distributions.von_mises_fisher import VonMisesFisher
 from torch.distributions.normal import Normal
 from torch.nn import functional as F
 

neurometry/curvature/models/toroidal_vae.py

@@ -5,7 +5,7 @@
 
 import geomstats.backend as gs
 import torch
-from hyperspherical.distributions import VonMisesFisher
+from neurometry.curvature.hyperspherical.distributions.von_mises_fisher import VonMisesFisher
 from torch.nn import functional as F
 
 

neurometry/curvature/train.py

@@ -2,7 +2,7 @@
 
 import copy
 
-import losses
+import neurometry.curvature.losses as losses
 import torch
 import wandb
 

neurometry/datasets/load_rnn_grid_cells.py

@@ -12,60 +12,31 @@
     utils,
 )
 
-# Loading single agent model
 
-# parent_dir = os.getcwd() + "/datasets/rnn_grid_cells/"
-
-parent_dir = "/scratch/facosta/rnn_grid_cells/"
-
-
-single_model_folder = "Single agent path integration/Seed 1 weight decay 1e-06/"
-single_model_parameters = "steps_20_batch_200_RNN_4096_relu_rf_012_DoG_True_periodic_False_lr_00001_weight_decay_1e-06/"
-
-
-dual_model_folder = (
-    "Dual agent path integration disjoint PCs/Seed 1 weight decay 1e-06/"
-)
-dual_model_parameters = "steps_20_batch_200_RNN_4096_relu_rf_012_DoG_True_periodic_False_lr_00001_weight_decay_1e-06/"
-
-
-def load_activations(epochs, version="single", verbose=True):
+def load_activations(epochs, file_path, version="single", verbose=True, save = True):
     activations = []
     rate_maps = []
     state_points = []
     positions = []
+    g_s = []
 
-    if version == "single":
-        activations_dir = (
-            parent_dir + single_model_folder + single_model_parameters + "activations/"
-        )
-    elif version == "dual":
-        activations_dir = (
-            parent_dir + dual_model_folder + dual_model_parameters + "activations/"
-        )
+    activations_dir = os.path.join(file_path, "activations")
 
-    random.seed(0)
     for epoch in epochs:
-        activations_epoch_path = (
-            activations_dir + f"activations_{version}_agent_epoch_{epoch}.npy"
-        )
-        rate_map_epoch_path = (
-            activations_dir + f"rate_map_{version}_agent_epoch_{epoch}.npy"
-        )
-        positions_epoch_path = (
-            activations_dir + f"positions_{version}_agent_epoch_{epoch}.npy"
-        )
-
-        if (
-            os.path.exists(activations_epoch_path)
-            and os.path.exists(rate_map_epoch_path)
-            and os.path.exists(positions_epoch_path)
-        ):
+        activations_epoch_path = os.path.join(activations_dir, f"activations_{version}_agent_epoch_{epoch}.npy")
+        rate_map_epoch_path = os.path.join(activations_dir, f"rate_map_{version}_agent_epoch_{epoch}.npy")
+        positions_epoch_path = os.path.join(activations_dir, f"positions_{version}_agent_epoch_{epoch}.npy")
+        gs_epoch_path = os.path.join(activations_dir, f"g_{version}_agent_epoch_{epoch}.npy")
+
+        if os.path.exists(activations_epoch_path) and os.path.exists(
+            rate_map_epoch_path
+        ) and os.path.exists(positions_epoch_path) and os.path.exists(gs_epoch_path):
             activations.append(np.load(activations_epoch_path))
             rate_maps.append(np.load(rate_map_epoch_path))
             positions.append(np.load(positions_epoch_path))
+            g_s.append(np.load(gs_epoch_path))
             if verbose:
-                print(f"Epoch {epoch} found!")
+                print(f"Epoch {epoch} found.")
         else:
             print(f"Epoch {epoch} not found. Loading ...")
             parser = config.parser
@@ -75,22 +46,32 @@ def load_activations(epochs, version="single", verbose=True):
                 (
                     activations_single_agent,
                     rate_map_single_agent,
+                    g_single_agent,
                     positions_single_agent,
-                ) = single_agent_activity.main(options, epoch=epoch)
+                ) = single_agent_activity.main(options, file_path, epoch=epoch)
                 activations.append(activations_single_agent)
                 rate_maps.append(rate_map_single_agent)
                 positions.append(positions_single_agent)
+                g_s.append(g_single_agent)
             elif version == "dual":
-                activations_dual_agent, rate_map_dual_agent, positions_dual_agent = (
-                    dual_agent_activity.main(options, epoch=epoch)
-                )
+                activations_dual_agent, rate_map_dual_agent, g_dual_agent, positions_dual_agent = dual_agent_activity.main(
+                    options, file_path, epoch=epoch)
                 activations.append(activations_dual_agent)
                 rate_maps.append(rate_map_dual_agent)
                 positions.append(positions_dual_agent)
-        print(len(activations))
+                g_s.append(g_dual_agent)
+
+            if save:
+                np.save(activations_epoch_path, activations[-1])
+                np.save(rate_map_epoch_path, rate_maps[-1])
+                np.save(positions_epoch_path, positions[-1])
+                np.save(gs_epoch_path, g_s[-1])
+
         state_points_epoch = activations[-1].reshape(activations[-1].shape[0], -1)
         state_points.append(state_points_epoch)
 
+
+
     if verbose:
         print(f"Loaded epochs {epochs} of {version} agent model.")
         print(
@@ -104,7 +85,7 @@ def load_activations(epochs, version="single", verbose=True):
         )
         print(f"positions has shape {positions[0].shape}.")
 
-    return activations, rate_maps, state_points, positions
+    return activations, rate_maps, state_points, positions, g_s
 
 
 # def plot_rate_map(indices, num_plots, activations, title):
@@ -137,9 +118,8 @@ def load_activations(epochs, version="single", verbose=True):
 #     plt.show()
 
 
-
-def plot_rate_map(indices, num_plots, activations, title):
-    rng = np.random.default_rng(seed=0)
+def plot_rate_map(indices, num_plots, activations, title, seed=None):
+    rng = np.random.default_rng(seed=seed)
     if indices is None:
         idxs = rng.integers(0, activations.shape[0] - 1, num_plots)
     else:

neurometry/datasets/rnn_grid_cells/config.py

@@ -28,6 +28,7 @@ class Config:
     n_avg = 50  # number of trajectories to average over for rate maps
 
 
+
 # If you need to access the configuration as a dictionary
 config = Config.__dict__