refactor: continue analysis/plot restructuring

Author: fabioseel

retinal_rl/analysis/channel_analysis.py

@@ -1,6 +1,3 @@
-
-
-
 import logging
 from dataclasses import dataclass
 from typing import cast
@@ -23,6 +20,7 @@
 
 logger = logging.getLogger(__name__)
 
+
 @dataclass
 class SpectralAnalysis:
     """Results of spectral analysis for a layer."""
@@ -40,6 +38,7 @@ class HistogramAnalysis:
     channel_histograms: FloatArray
     bin_edges: FloatArray
 
+
 def spectral_analysis(
     device: torch.device,
     imageset: Imageset,
@@ -54,7 +53,7 @@ def spectral_analysis(
     dataloader = _prepare_dataset(imageset, max_sample_size)
 
     # Initialize results
-    results = {"input": _layer_spectral_analysis(device, dataloader, nn.Identity())}
+    results: dict[str, SpectralAnalysis] = {}
 
     # Analyze each layer
     head_layers: list[nn.Module] = []
@@ -64,7 +63,6 @@ def spectral_analysis(
 
         if is_nonlinearity(layer):
             continue
-        # TODO: Possible for non Conv2D layers?
 
         results[layer_name] = _layer_spectral_analysis(
             device, dataloader, nn.Sequential(*head_layers)
@@ -84,10 +82,10 @@ def histogram_analysis(
     _, cnn_layers = get_cnn_circuit(brain)
 
     # Prepare dataset
-    dataloader = _prepare_dataset(imageset, max_sample_size)
+    dataloader = _prepare_dataset(imageset, max_sample_size)  # TODO: Move outside?
 
     # Initialize results
-    results = {"input": _layer_pixel_histograms(device, dataloader, nn.Identity())}
+    results: dict[str, HistogramAnalysis] = {}
 
     # Analyze each layer
     head_layers: list[nn.Module] = []
@@ -96,7 +94,6 @@ def histogram_analysis(
         head_layers.append(layer)
         if is_nonlinearity(layer):
             continue
-        # TODO: Possible for non Conv2D layers?
         results[layer_name] = _layer_pixel_histograms(
             device, dataloader, nn.Sequential(*head_layers)
         )
@@ -243,8 +240,6 @@ def plot(
     copy_checkpoint: bool,
 ):
     for layer_name, layer_rfs in rf_result.items():
-        if layer_name != "input":
-            continue
         layer_spectral = spectral_result[layer_name]
         layer_histogram = histogram_result[layer_name]
         for channel in range(layer_rfs.shape[0]):
@@ -395,4 +390,37 @@ def _plot_receptive_fields(ax: Axes, rf: FloatArray):
         horizontalalignment="center",
         verticalalignment="center",
         transform=ax.transAxes,
-    )
+    )
+
+
+def analyze_input(
+    device: torch.device, imageset: Imageset, max_sample_size: int
+) -> tuple[SpectralAnalysis, HistogramAnalysis]:
+    dataloader = _prepare_dataset(imageset, max_sample_size)
+    spectral_result = _layer_spectral_analysis(device, dataloader, nn.Identity())
+    histogram_result = _layer_pixel_histograms(device, dataloader, nn.Identity())
+    return spectral_result, histogram_result
+
+
+def input_plot(
+    log: FigureLogger,
+    rf_result: FloatArray,
+    spectral_result: SpectralAnalysis,
+    histogram_result: HistogramAnalysis,
+    init_dir: str,
+):
+    for channel in range(histogram_result.channel_histograms.shape[0]):
+        channel_fig = layer_channel_plots(
+            rf_result,
+            spectral_result,
+            histogram_result,
+            layer_name="input",
+            channel=channel,
+        )
+        log.log_figure(
+            channel_fig,
+            init_dir,
+            f"input_channel_{channel}",
+            0,
+            False,
+        )

retinal_rl/analysis/default.py

@@ -0,0 +1,38 @@
+from retinal_rl.analysis.plot import (
+    FigureLogger,
+    plot_brain_and_optimizers,
+    plot_receptive_field_sizes,
+)
+from retinal_rl.models.brain import Brain
+from retinal_rl.models.objective import ContextT, Objective
+from retinal_rl.util import FloatArray
+
+INIT_DIR = "initialization_analysis"
+
+def initialization_plots(
+    log: FigureLogger,
+    brain: Brain,
+    objective: Objective[ContextT],
+    input_shape: tuple[int, ...],
+    rf_result: dict[str, FloatArray],
+):
+    log.save_summary(brain)
+
+    # TODO: This is a bit of a hack, we should refactor this to get the relevant information out of  cnn_stats
+    rf_sizes_fig = plot_receptive_field_sizes(input_shape, rf_result)
+    log.log_figure(
+        rf_sizes_fig,
+        INIT_DIR,
+        "receptive_field_sizes",
+        0,
+        False,
+    )
+
+    graph_fig = plot_brain_and_optimizers(brain, objective)
+    log.log_figure(
+        graph_fig,
+        INIT_DIR,
+        "brain_graph",
+        0,
+        False,
+    )

retinal_rl/analysis/plot.py

@@ -1,7 +1,9 @@
 """Utility functions for plotting the results of statistical analyses."""
 
+import json
 import shutil
 from pathlib import Path
+from typing import Any
 
 import matplotlib.pyplot as plt
 import networkx as nx
@@ -17,7 +19,7 @@
 
 from retinal_rl.models.brain import Brain
 from retinal_rl.models.objective import ContextT, Objective
-from retinal_rl.util import FloatArray
+from retinal_rl.util import FloatArray, NumpyEncoder
 
 
 def make_image_grid(arrays: list[FloatArray], nrow: int) -> FloatArray:
@@ -178,8 +180,6 @@ def plot_receptive_field_sizes(
     rf_sizes: list[tuple[int, int]] = []
     layer_names: list[str] = []
     for name, rf in rf_layers.items():
-        if name == "input":  # TODO: Should not be possible?!
-            continue
         rf_height, rf_width = rf.shape[2:]
         rf_sizes.append((rf_height, rf_width))
         layer_names.append(name)
@@ -291,13 +291,10 @@ def set_integer_ticks(ax: Axes):
     ax.xaxis.set_major_locator(MaxNLocator(integer=True))
     ax.yaxis.set_major_locator(MaxNLocator(integer=True))
 
+
 class FigureLogger:
     def __init__(
-        self,
-        use_wandb: bool,
-        plot_dir: Path,
-        checkpoint_plot_dir: Path,
-        run_dir: Path
+        self, use_wandb: bool, plot_dir: Path, checkpoint_plot_dir: Path, run_dir: Path
     ):
         self.use_wandb = use_wandb
         self.plot_dir = plot_dir
@@ -340,7 +337,6 @@ def capitalize_part(part: str) -> str:
         return "/".join(capitalized_parts)
 
     def _checkpoint_copy(self, sub_dir: str, file_name: str, epoch: int) -> None:
-        # TODO: Does this need to be in here?
         src_path = self.plot_dir / sub_dir / f"{file_name}.png"
 
         dest_dir = self.checkpoint_plot_dir / f"epoch_{epoch}" / sub_dir
@@ -370,3 +366,7 @@ def save_summary(self, brain: Brain):
 
         if self.use_wandb:
             wandb.save(str(filepath), base_path=self.run_dir, policy="now")
+
+    def save_dict(self, path: Path, dict: dict[str, Any]):
+        with open(path, "w") as f:
+            json.dump(dict, f, cls=NumpyEncoder)

retinal_rl/analysis/receptive_fields.py

@@ -40,15 +40,14 @@ def analyze(brain: Brain, device: torch.device):
 
     return input_shape, results
 
+
 def plot(
     log: FigureLogger,
     rf_result: dict[str, FloatArray],
     epoch: int,
     copy_checkpoint: bool,
 ):
     for layer_name, layer_rfs in rf_result.items():
-        if layer_name == "input": # TODO: Remove, see where this goes
-            continue
         layer_rf_plots = layer_receptive_field_plots(layer_rfs)
         log.log_figure(
             layer_rf_plots,
@@ -58,9 +57,8 @@ def plot(
             copy_checkpoint,
         )
 
-def layer_receptive_field_plots(
-    lyr_rfs: FloatArray, max_cols: int = 8
-) -> Figure:
+
+def layer_receptive_field_plots(lyr_rfs: FloatArray, max_cols: int = 8) -> Figure:
     """Plot the receptive fields of a convolutional layer."""
     ochns, _, _, _ = lyr_rfs.shape
 
@@ -79,9 +77,7 @@ def layer_receptive_field_plots(
 
     for i in range(ochns):
         ax = axs[i]
-        data = np.moveaxis(
-            lyr_rfs[i], 0, -1
-        )  # Move channel axis to the last dimension
+        data = np.moveaxis(lyr_rfs[i], 0, -1)  # Move channel axis to the last dimension
         data_min = data.min()
         data_max = data.max()
         data = (data - data_min) / (data_max - data_min)
@@ -96,6 +92,7 @@ def layer_receptive_field_plots(
     fig.tight_layout()  # Adjust layout to fit color bars
     return fig
 
+
 def _compute_receptive_fields(
     device: torch.device,
     head_layers: list[nn.Module],

retinal_rl/analysis/reconstructions.py

@@ -1,5 +1,3 @@
-
-import json
 from dataclasses import asdict, dataclass
 from pathlib import Path
 
@@ -13,7 +11,7 @@
 from retinal_rl.models.brain import Brain
 from retinal_rl.models.loss import ContextT, ReconstructionLoss
 from retinal_rl.models.objective import Objective
-from retinal_rl.util import FloatArray, NumpyEncoder
+from retinal_rl.util import FloatArray
 
 
 @dataclass
@@ -24,42 +22,47 @@ class Reconstructions:
     inputs: list[tuple[FloatArray, int]]
     estimates: list[tuple[FloatArray, int]]
 
+
 @dataclass
 class ReconstructionStatistics:
     """Results of image reconstruction for both training and test sets."""
 
     train: Reconstructions
     test: Reconstructions
 
-# TODO: Make structure match the analyze / plot structure as receptive_fields
 
-def perform_reconstruction_analysis(
-    log: FigureLogger,
-    analyses_dir: Path,
+def analyze(
     device: torch.device,
     brain: Brain,
     objective: Objective[ContextT],
     train_set: Imageset,
     test_set: Imageset,
-    epoch: int,
-    copy_checkpoint: bool,
-):
+) -> tuple[dict[str, ReconstructionStatistics], list[float], list[float]]:
     reconstruction_decoders = [
         loss.target_decoder
         for loss in objective.losses
         if isinstance(loss, ReconstructionLoss)
     ]
 
+    results: dict[str, ReconstructionStatistics] = {}
     for decoder in reconstruction_decoders:
-        norm_means, norm_stds = train_set.normalization_stats
-        rec_dict = asdict(
-            reconstruct_images(device, brain, decoder, train_set, test_set, 5)
+        results[decoder] = reconstruct_images(
+            device, brain, decoder, train_set, test_set, 5
         )
-        # Save the reconstructions
-        rec_path = analyses_dir / f"{decoder}_reconstructions_epoch_{epoch}.json"
-        with open(rec_path, "w") as f:
-            json.dump(rec_dict, f, cls=NumpyEncoder)
+    return results, *train_set.normalization_stats
 
+
+def plot(
+    log: FigureLogger,
+    analyses_dir: Path,
+    result: dict[str, ReconstructionStatistics],
+    norm_means: list[float],
+    norm_stds: list[float],
+    epoch: int,
+    copy_checkpoint: bool,
+):
+    for decoder, reconstructions in result.items():
+        rec_dict = asdict(reconstructions)
         recon_fig = plot_reconstructions(
             norm_means,
             norm_stds,
@@ -74,6 +77,10 @@ def perform_reconstruction_analysis(
             epoch,
             copy_checkpoint,
         )
+        # Save the reconstructions #TODO: most plot functions don't do this, should stay?
+        log.save_dict(
+            analyses_dir / f"{decoder}_reconstructions_epoch_{epoch}.json", rec_dict
+        )
 
 
 def reconstruct_images(

retinal_rl/analysis/transforms_analysis.py

@@ -18,11 +18,8 @@ class TransformStatistics:
     source_transforms: dict[str, dict[float, list[FloatArray]]]
     noise_transforms: dict[str, dict[float, list[FloatArray]]]
 
-# TODO: Make structure match the analyze / plot structure as receptive_fields
 
-def transform_base_images(
-    imageset: Imageset, num_steps: int, num_images: int
-) -> TransformStatistics:
+def analyze(imageset: Imageset, num_steps: int, num_images: int) -> TransformStatistics:
     """Apply transformations to a set of images from an Imageset."""
     images: list[Image.Image] = []
 
@@ -59,7 +56,7 @@ def transform_base_images(
     return resultss
 
 
-def plot_transforms(
+def plot(
     source_transforms: dict[str, dict[float, list[FloatArray]]],
     noise_transforms: dict[str, dict[float, list[FloatArray]]],
 ) -> Figure:
@@ -142,4 +139,4 @@ def plot_transforms(
         transform_index += 1
 
     plt.tight_layout()
-    return fig
+    return fig

retinal_rl/util.py

@@ -19,6 +19,7 @@
 
 ### IO Handling stuff
 
+
 class NumpyEncoder(json.JSONEncoder):
     """JSON encoder that handles numpy arrays."""
 
@@ -27,6 +28,7 @@ def default(self, obj: Any) -> Any:
             return obj.tolist()
         return super().default(obj)
 
+
 ### Functions