Implement 8bit numbers

configs/data.yaml

@@ -95,6 +95,22 @@ straightline:
     num_aug_versions: 0
     noise: 0.0
 
+eight_bit_numbers:
+  num_channels: 1
+  mean: [ 0. ]
+  std: [ 1. ]
+  img_width: 32
+  img_height: 32
+  train_dataset_params:
+    samples_per_class: 50
+    include_noise: False
+  valid_dataset_params:
+    samples_per_class: 1
+    include_noise: False
+  test_dataset_params:
+    samples_per_class: 1
+    include_noise: False
+
 augmentation_v1:
   aug1:
     - probability: 0.8

configs/s1_toy_sample2.yaml

@@ -0,0 +1,76 @@
+dataset:
+  name: eight_bit_numbers
+  augmentation: None
+  loader: torch
+  batch_size: 1
+  num_workers: 0
+
+run:
+  n_epochs: 15
+  current_epoch: 0
+  plots:
+    enable: True
+    only_last_epoch: False
+    store_path: "../tmp/test/"
+  store_state_path: None
+  load_state_path: None
+
+
+metrics:
+  - mse:
+      type: MeanSquaredError
+      meter: avg
+  - mae:
+      type: MeanAbsoluteError
+      meter: avg
+
+optimizers:
+  l2_opt:
+    type: Adam
+    params:
+      lr: 0.05
+      betas: [ 0.9, 0.999 ]
+      eps: 0.00000001
+      weight_decay: 0
+      amsgrad: False
+    scheduler:
+      type: ReduceLROnPlateau
+      params:
+        mode: min
+        factor: 0.1
+        patience: 5
+        threshold: 0.0001
+        threshold_mode: rel
+        cooldown: 0
+        min_lr: 0.000001
+        eps: 0.00000001
+
+logging:
+  wandb:
+    active: False,
+    save_dir: "../wandb"
+    project: "lateral_connections_toy_example"
+    log_model: True
+    group: null # "hebbian_learning"
+    job_type: null # "train"
+  console:
+    active: True
+
+feature_extractor:
+  out_channels: 4
+  add_bg_channel: False
+  bin_threshold: 0.  # set to 0.5 to obtain better features
+  optimized_filter_lines: True # set to True to obtain better features
+
+lateral_model:
+  channels: 4
+  max_timesteps: 6  # 1 = only one forward pass without recurrent connection
+  min_k: 2
+  max_k: 3
+  l1_type: 'lateral_flex'
+  l1_params:
+    locality_size: 5
+    lr: 0.2
+    hebbian_rule: 'vanilla'
+    neg_corr: True
+    act_threshold: 0.4  # 'bernoulli'

src/changing_line_demo.py

@@ -159,10 +159,10 @@ def create_image(self, img: Tensor, in_features: Tensor, l1_act: Tensor, l2_act:
         l2_act = Image.fromarray(self.to_mask((l2_act * 255).squeeze().cpu().numpy()))
 
         # Resize Images
-        img = img.resize((self.img_size, self.img_size), Image.Resampling.LANCZOS)
-        in_features = in_features.resize((self.img_size, self.img_size), Image.Resampling.LANCZOS)
-        l1_act = l1_act.resize((self.img_size, self.img_size), Image.Resampling.LANCZOS)
-        l2_act = l2_act.resize((self.img_size, self.img_size), Image.Resampling.LANCZOS)
+        img = img.resize((self.img_size, self.img_size), Image.Resampling.NEAREST)
+        in_features = in_features.resize((self.img_size, self.img_size), Image.Resampling.NEAREST)
+        l1_act = l1_act.resize((self.img_size, self.img_size), Image.Resampling.NEAREST)
+        l2_act = l2_act.resize((self.img_size, self.img_size), Image.Resampling.NEAREST)
 
         output = self.img_template.copy()