diff --git a/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom.yaml b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom.yaml
new file mode 100644
index 000000000..835847ba1
--- /dev/null
+++ b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom.yaml
@@ -0,0 +1,114 @@
+---
+trainer:
+  class_path: eva.Trainer
+  init_args:
+    n_runs: &N_RUNS ${oc.env:N_RUNS, 5}
+    default_root_dir: &OUTPUT_ROOT ${oc.env:OUTPUT_ROOT, logs/${oc.env:MODEL_NAME, dino_vits16}/offline/bach}
+    max_steps: &MAX_STEPS ${oc.env:MAX_STEPS, 6000} #usually 12500
+    checkpoint_type: ${oc.env:CHECKPOINT_TYPE, best}
+    callbacks:
+      - class_path: eva.callbacks.ConfigurationLogger
+      - class_path: lightning.pytorch.callbacks.TQDMProgressBar
+        init_args:
+          refresh_rate: ${oc.env:TQDM_REFRESH_RATE, 1}
+      - class_path: lightning.pytorch.callbacks.LearningRateMonitor
+        init_args:
+          logging_interval: epoch
+      - class_path: lightning.pytorch.callbacks.ModelCheckpoint
+        init_args:
+          filename: best
+          save_last: ${oc.env:SAVE_LAST, false}
+          save_top_k: 1
+          monitor: &MONITOR_METRIC ${oc.env:MONITOR_METRIC, val/MulticlassAccuracy}
+          mode: &MONITOR_METRIC_MODE ${oc.env:MONITOR_METRIC_MODE, max}
+      - class_path: lightning.pytorch.callbacks.EarlyStopping
+        init_args:
+          min_delta: 0
+          patience: ${oc.env:PATIENCE, 1250}
+          monitor: *MONITOR_METRIC
+          mode: *MONITOR_METRIC_MODE
+      - class_path: eva.callbacks.ClassificationEmbeddingsWriter
+        init_args:
+          output_dir: &DATASET_EMBEDDINGS_ROOT ${oc.env:EMBEDDINGS_ROOT, ./data/embeddings}/${oc.env:MODEL_NAME, dino_vits16}/bach
+          dataloader_idx_map:
+            0: train
+            1: val
+          backbone:
+            class_path: eva.core.models.wrappers.ModelFromLocal
+            init_args:
+              local_repo_path: "./dinov3"
+              model_name: "dinov3_vith16plus" #The model you want to load
+              checkpoint_path: "./dinov3_vith16plus_pretrain_lvd1689m-7c1da9a5.pth" #Path to the model checkpoint
+ 
+
+          overwrite: true
+    logger:
+      - class_path: lightning.pytorch.loggers.TensorBoardLogger
+        init_args:
+          save_dir: *OUTPUT_ROOT
+          name: ""
+model:
+  class_path: eva.HeadModule
+  init_args:
+    head:
+      class_path: eva.core.models.networks.batch_linear.batch_linear
+      init_args:
+        in_features: ${oc.env:IN_FEATURES, 1280} #MODIFY ME
+        out_features: &NUM_CLASSES 4
+    criterion: torch.nn.CrossEntropyLoss
+    optimizer:
+      class_path: torch.optim.AdamW
+      init_args:
+        lr: ${oc.env:LR_VALUE, 0.0001}
+    metrics:
+      common:
+        - class_path: eva.metrics.AverageLoss
+        - class_path: eva.metrics.MulticlassClassificationMetrics
+          init_args:
+            num_classes: *NUM_CLASSES
+data:
+  class_path: eva.DataModule
+  init_args:
+    datasets:
+      train:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args: &DATASET_ARGS
+          root: *DATASET_EMBEDDINGS_ROOT
+          manifest_file: manifest.csv
+          split: train
+      val:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args:
+          <<: *DATASET_ARGS
+          split: val
+      predict:
+        - class_path: eva.vision.datasets.BACH
+          init_args: &PREDICT_DATASET_ARGS
+            root: ${oc.env:DATA_ROOT, ./data/bach}
+            split: train
+            download: ${oc.env:DOWNLOAD_DATA, false}
+            # Set `download: true` to download the dataset from https://zenodo.org/records/3632035
+            # The BACH dataset is distributed under the following license
+            # Attribution-NonCommercial-NoDerivs 4.0 International license
+            # (see: https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode)
+            transforms:
+              class_path: eva.vision.data.transforms.common.ResizeAndCrop
+              init_args:
+                size: ${oc.env:RESIZE_DIM, 224}  
+                mean: ${oc.env:NORMALIZE_MEAN, [0.485, 0.456, 0.406]} #Flagged for maybe needs fixing
+                std: ${oc.env:NORMALIZE_STD, [0.229, 0.224, 0.225]}
+        - class_path: eva.vision.datasets.BACH
+          init_args:
+            <<: *PREDICT_DATASET_ARGS
+            split: val
+    dataloaders:
+      train:
+        batch_size: &BATCH_SIZE ${oc.env:BATCH_SIZE, 256}
+        num_workers: &N_DATA_WORKERS ${oc.env:N_DATA_WORKERS, 4}
+        shuffle: true
+      val:
+        batch_size: *BATCH_SIZE
+        num_workers: *N_DATA_WORKERS
+      predict:
+        batch_size: &PREDICT_BATCH_SIZE ${oc.env:PREDICT_BATCH_SIZE, 64}
+        num_workers: *N_DATA_WORKERS
diff --git a/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear.yaml b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear.yaml
new file mode 100644
index 000000000..af46c4d2e
--- /dev/null
+++ b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear.yaml
@@ -0,0 +1,114 @@
+---
+trainer:
+  class_path: eva.Trainer
+  init_args:
+    n_runs: &N_RUNS ${oc.env:N_RUNS, 5}
+    default_root_dir: &OUTPUT_ROOT ${oc.env:OUTPUT_ROOT, logs/${oc.env:MODEL_NAME, dino_vits16}/offline/bach}
+    max_steps: &MAX_STEPS ${oc.env:MAX_STEPS, 6000} #usually 12500
+    checkpoint_type: ${oc.env:CHECKPOINT_TYPE, best}
+    callbacks:
+      - class_path: eva.callbacks.ConfigurationLogger
+      - class_path: lightning.pytorch.callbacks.TQDMProgressBar
+        init_args:
+          refresh_rate: ${oc.env:TQDM_REFRESH_RATE, 1}
+      - class_path: lightning.pytorch.callbacks.LearningRateMonitor
+        init_args:
+          logging_interval: epoch
+      - class_path: lightning.pytorch.callbacks.ModelCheckpoint
+        init_args:
+          filename: best
+          save_last: ${oc.env:SAVE_LAST, false}
+          save_top_k: 1
+          monitor: &MONITOR_METRIC ${oc.env:MONITOR_METRIC, val/MulticlassAccuracy}
+          mode: &MONITOR_METRIC_MODE ${oc.env:MONITOR_METRIC_MODE, max}
+      - class_path: lightning.pytorch.callbacks.EarlyStopping
+        init_args:
+          min_delta: 0
+          patience: ${oc.env:PATIENCE, 1250}
+          monitor: *MONITOR_METRIC
+          mode: *MONITOR_METRIC_MODE
+      - class_path: eva.callbacks.ClassificationEmbeddingsWriter
+        init_args:
+          output_dir: &DATASET_EMBEDDINGS_ROOT ${oc.env:EMBEDDINGS_ROOT, ./data/embeddings}/${oc.env:MODEL_NAME, dino_vits16}/bach
+          dataloader_idx_map:
+            0: train
+            1: val
+          backbone:
+            class_path: eva.core.models.wrappers.ModelFromLocal
+            init_args:
+              local_repo_path: "./dinov3"
+              model_name: "dinov3_vith16plus" #The model you want to load
+              checkpoint_path: "./dinov3_vith16plus_pretrain_lvd1689m-7c1da9a5.pth" #Path to the model checkpoint
+ 
+
+          overwrite: true
+    logger:
+      - class_path: lightning.pytorch.loggers.TensorBoardLogger
+        init_args:
+          save_dir: *OUTPUT_ROOT
+          name: ""
+model:
+  class_path: eva.HeadModule
+  init_args:
+    head:
+      class_path: eva.core.models.networks.adj_pooled_linear.adj_pooled_linear
+      init_args:
+        in_features: ${oc.env:IN_FEATURES, 1280} #MODIFY ME
+        out_features: &NUM_CLASSES 4
+    criterion: torch.nn.CrossEntropyLoss
+    optimizer:
+      class_path: torch.optim.AdamW
+      init_args:
+        lr: ${oc.env:LR_VALUE, 0.0001}
+    metrics:
+      common:
+        - class_path: eva.metrics.AverageLoss
+        - class_path: eva.metrics.MulticlassClassificationMetrics
+          init_args:
+            num_classes: *NUM_CLASSES
+data:
+  class_path: eva.DataModule
+  init_args:
+    datasets:
+      train:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args: &DATASET_ARGS
+          root: *DATASET_EMBEDDINGS_ROOT
+          manifest_file: manifest.csv
+          split: train
+      val:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args:
+          <<: *DATASET_ARGS
+          split: val
+      predict:
+        - class_path: eva.vision.datasets.BACH
+          init_args: &PREDICT_DATASET_ARGS
+            root: ${oc.env:DATA_ROOT, ./data/bach}
+            split: train
+            download: ${oc.env:DOWNLOAD_DATA, false}
+            # Set `download: true` to download the dataset from https://zenodo.org/records/3632035
+            # The BACH dataset is distributed under the following license
+            # Attribution-NonCommercial-NoDerivs 4.0 International license
+            # (see: https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode)
+            transforms:
+              class_path: eva.vision.data.transforms.common.ResizeAndCrop
+              init_args:
+                size: ${oc.env:RESIZE_DIM, 224}  
+                mean: ${oc.env:NORMALIZE_MEAN, [0.485, 0.456, 0.406]} #Flagged for maybe needs fixing
+                std: ${oc.env:NORMALIZE_STD, [0.229, 0.224, 0.225]}
+        - class_path: eva.vision.datasets.BACH
+          init_args:
+            <<: *PREDICT_DATASET_ARGS
+            split: val
+    dataloaders:
+      train:
+        batch_size: &BATCH_SIZE ${oc.env:BATCH_SIZE, 256}
+        num_workers: &N_DATA_WORKERS ${oc.env:N_DATA_WORKERS, 4}
+        shuffle: true
+      val:
+        batch_size: *BATCH_SIZE
+        num_workers: *N_DATA_WORKERS
+      predict:
+        batch_size: &PREDICT_BATCH_SIZE ${oc.env:PREDICT_BATCH_SIZE, 64}
+        num_workers: *N_DATA_WORKERS
diff --git a/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear_noreg.yaml b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear_noreg.yaml
new file mode 100644
index 000000000..65d5c1817
--- /dev/null
+++ b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_adj_pooled_linear_noreg.yaml
@@ -0,0 +1,114 @@
+---
+trainer:
+  class_path: eva.Trainer
+  init_args:
+    n_runs: &N_RUNS ${oc.env:N_RUNS, 5}
+    default_root_dir: &OUTPUT_ROOT ${oc.env:OUTPUT_ROOT, logs/${oc.env:MODEL_NAME, dino_vits16}/offline/bach}
+    max_steps: &MAX_STEPS ${oc.env:MAX_STEPS, 6000} #usually 12500
+    checkpoint_type: ${oc.env:CHECKPOINT_TYPE, best}
+    callbacks:
+      - class_path: eva.callbacks.ConfigurationLogger
+      - class_path: lightning.pytorch.callbacks.TQDMProgressBar
+        init_args:
+          refresh_rate: ${oc.env:TQDM_REFRESH_RATE, 1}
+      - class_path: lightning.pytorch.callbacks.LearningRateMonitor
+        init_args:
+          logging_interval: epoch
+      - class_path: lightning.pytorch.callbacks.ModelCheckpoint
+        init_args:
+          filename: best
+          save_last: ${oc.env:SAVE_LAST, false}
+          save_top_k: 1
+          monitor: &MONITOR_METRIC ${oc.env:MONITOR_METRIC, val/MulticlassAccuracy}
+          mode: &MONITOR_METRIC_MODE ${oc.env:MONITOR_METRIC_MODE, max}
+      - class_path: lightning.pytorch.callbacks.EarlyStopping
+        init_args:
+          min_delta: 0
+          patience: ${oc.env:PATIENCE, 1250}
+          monitor: *MONITOR_METRIC
+          mode: *MONITOR_METRIC_MODE
+      - class_path: eva.callbacks.ClassificationEmbeddingsWriter
+        init_args:
+          output_dir: &DATASET_EMBEDDINGS_ROOT ${oc.env:EMBEDDINGS_ROOT, ./data/embeddings}/${oc.env:MODEL_NAME, dino_vits16}/bach
+          dataloader_idx_map:
+            0: train
+            1: val
+          backbone:
+            class_path: eva.core.models.wrappers.ModelFromLocal
+            init_args:
+              local_repo_path: "./dinov3"
+              model_name: "dinov3_vith16plus" #The model you want to load
+              checkpoint_path: "./dinov3_vith16plus_pretrain_lvd1689m-7c1da9a5.pth" #Path to the model checkpoint
+ 
+
+          overwrite: true
+    logger:
+      - class_path: lightning.pytorch.loggers.TensorBoardLogger
+        init_args:
+          save_dir: *OUTPUT_ROOT
+          name: ""
+model:
+  class_path: eva.HeadModule
+  init_args:
+    head:
+      class_path: eva.core.models.networks.adj_pooled_linear_noreg.adj_pooled_linear_noreg
+      init_args:
+        in_features: ${oc.env:IN_FEATURES, 1280} #MODIFY ME
+        out_features: &NUM_CLASSES 4
+    criterion: torch.nn.CrossEntropyLoss
+    optimizer:
+      class_path: torch.optim.AdamW
+      init_args:
+        lr: ${oc.env:LR_VALUE, 0.0001}
+    metrics:
+      common:
+        - class_path: eva.metrics.AverageLoss
+        - class_path: eva.metrics.MulticlassClassificationMetrics
+          init_args:
+            num_classes: *NUM_CLASSES
+data:
+  class_path: eva.DataModule
+  init_args:
+    datasets:
+      train:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args: &DATASET_ARGS
+          root: *DATASET_EMBEDDINGS_ROOT
+          manifest_file: manifest.csv
+          split: train
+      val:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args:
+          <<: *DATASET_ARGS
+          split: val
+      predict:
+        - class_path: eva.vision.datasets.BACH
+          init_args: &PREDICT_DATASET_ARGS
+            root: ${oc.env:DATA_ROOT, ./data/bach}
+            split: train
+            download: ${oc.env:DOWNLOAD_DATA, false}
+            # Set `download: true` to download the dataset from https://zenodo.org/records/3632035
+            # The BACH dataset is distributed under the following license
+            # Attribution-NonCommercial-NoDerivs 4.0 International license
+            # (see: https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode)
+            transforms:
+              class_path: eva.vision.data.transforms.common.ResizeAndCrop
+              init_args:
+                size: ${oc.env:RESIZE_DIM, 224}  
+                mean: ${oc.env:NORMALIZE_MEAN, [0.485, 0.456, 0.406]} #Flagged for maybe needs fixing
+                std: ${oc.env:NORMALIZE_STD, [0.229, 0.224, 0.225]}
+        - class_path: eva.vision.datasets.BACH
+          init_args:
+            <<: *PREDICT_DATASET_ARGS
+            split: val
+    dataloaders:
+      train:
+        batch_size: &BATCH_SIZE ${oc.env:BATCH_SIZE, 256}
+        num_workers: &N_DATA_WORKERS ${oc.env:N_DATA_WORKERS, 4}
+        shuffle: true
+      val:
+        batch_size: *BATCH_SIZE
+        num_workers: *N_DATA_WORKERS
+      predict:
+        batch_size: &PREDICT_BATCH_SIZE ${oc.env:PREDICT_BATCH_SIZE, 64}
+        num_workers: *N_DATA_WORKERS
diff --git a/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_tied_double_linear.yaml b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_tied_double_linear.yaml
new file mode 100644
index 000000000..863016308
--- /dev/null
+++ b/configs/vision/pathology/offline/classification/bach-linear-experiments/eval_dinov3_custom_tied_double_linear.yaml
@@ -0,0 +1,114 @@
+---
+trainer:
+  class_path: eva.Trainer
+  init_args:
+    n_runs: &N_RUNS ${oc.env:N_RUNS, 5}
+    default_root_dir: &OUTPUT_ROOT ${oc.env:OUTPUT_ROOT, logs/${oc.env:MODEL_NAME, dino_vits16}/offline/bach}
+    max_steps: &MAX_STEPS ${oc.env:MAX_STEPS, 6000} #usually 12500
+    checkpoint_type: ${oc.env:CHECKPOINT_TYPE, best}
+    callbacks:
+      - class_path: eva.callbacks.ConfigurationLogger
+      - class_path: lightning.pytorch.callbacks.TQDMProgressBar
+        init_args:
+          refresh_rate: ${oc.env:TQDM_REFRESH_RATE, 1}
+      - class_path: lightning.pytorch.callbacks.LearningRateMonitor
+        init_args:
+          logging_interval: epoch
+      - class_path: lightning.pytorch.callbacks.ModelCheckpoint
+        init_args:
+          filename: best
+          save_last: ${oc.env:SAVE_LAST, false}
+          save_top_k: 1
+          monitor: &MONITOR_METRIC ${oc.env:MONITOR_METRIC, val/MulticlassAccuracy}
+          mode: &MONITOR_METRIC_MODE ${oc.env:MONITOR_METRIC_MODE, max}
+      - class_path: lightning.pytorch.callbacks.EarlyStopping
+        init_args:
+          min_delta: 0
+          patience: ${oc.env:PATIENCE, 1250}
+          monitor: *MONITOR_METRIC
+          mode: *MONITOR_METRIC_MODE
+      - class_path: eva.callbacks.ClassificationEmbeddingsWriter
+        init_args:
+          output_dir: &DATASET_EMBEDDINGS_ROOT ${oc.env:EMBEDDINGS_ROOT, ./data/embeddings}/${oc.env:MODEL_NAME, dino_vits16}/bach
+          dataloader_idx_map:
+            0: train
+            1: val
+          backbone:
+            class_path: eva.core.models.wrappers.ModelFromLocal
+            init_args:
+              local_repo_path: "./dinov3"
+              model_name: "dinov3_vith16plus" #The model you want to load
+              checkpoint_path: "./dinov3_vith16plus_pretrain_lvd1689m-7c1da9a5.pth" #Path to the model checkpoint
+ 
+
+          overwrite: true
+    logger:
+      - class_path: lightning.pytorch.loggers.TensorBoardLogger
+        init_args:
+          save_dir: *OUTPUT_ROOT
+          name: ""
+model:
+  class_path: eva.HeadModule
+  init_args:
+    head:
+      class_path: eva.core.models.networks.tied_double_linear.tied_double_linear
+      init_args:
+        in_features: ${oc.env:IN_FEATURES, 1280} #MODIFY ME
+        out_features: &NUM_CLASSES 4
+    criterion: torch.nn.CrossEntropyLoss
+    optimizer:
+      class_path: torch.optim.AdamW
+      init_args:
+        lr: ${oc.env:LR_VALUE, 0.0001}
+    metrics:
+      common:
+        - class_path: eva.metrics.AverageLoss
+        - class_path: eva.metrics.MulticlassClassificationMetrics
+          init_args:
+            num_classes: *NUM_CLASSES
+data:
+  class_path: eva.DataModule
+  init_args:
+    datasets:
+      train:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args: &DATASET_ARGS
+          root: *DATASET_EMBEDDINGS_ROOT
+          manifest_file: manifest.csv
+          split: train
+      val:
+        class_path: eva.datasets.EmbeddingsClassificationDataset
+        init_args:
+          <<: *DATASET_ARGS
+          split: val
+      predict:
+        - class_path: eva.vision.datasets.BACH
+          init_args: &PREDICT_DATASET_ARGS
+            root: ${oc.env:DATA_ROOT, ./data/bach}
+            split: train
+            download: ${oc.env:DOWNLOAD_DATA, false}
+            # Set `download: true` to download the dataset from https://zenodo.org/records/3632035
+            # The BACH dataset is distributed under the following license
+            # Attribution-NonCommercial-NoDerivs 4.0 International license
+            # (see: https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode)
+            transforms:
+              class_path: eva.vision.data.transforms.common.ResizeAndCrop
+              init_args:
+                size: ${oc.env:RESIZE_DIM, 224}  
+                mean: ${oc.env:NORMALIZE_MEAN, [0.485, 0.456, 0.406]} #Flagged for maybe needs fixing
+                std: ${oc.env:NORMALIZE_STD, [0.229, 0.224, 0.225]}
+        - class_path: eva.vision.datasets.BACH
+          init_args:
+            <<: *PREDICT_DATASET_ARGS
+            split: val
+    dataloaders:
+      train:
+        batch_size: &BATCH_SIZE ${oc.env:BATCH_SIZE, 256}
+        num_workers: &N_DATA_WORKERS ${oc.env:N_DATA_WORKERS, 4}
+        shuffle: true
+      val:
+        batch_size: *BATCH_SIZE
+        num_workers: *N_DATA_WORKERS
+      predict:
+        batch_size: &PREDICT_BATCH_SIZE ${oc.env:PREDICT_BATCH_SIZE, 64}
+        num_workers: *N_DATA_WORKERS
diff --git a/eval_dinov3_custom.yaml b/eval_dinov3_custom.yaml
index d7a700fa2..835847ba1 100644
--- a/eval_dinov3_custom.yaml
+++ b/eval_dinov3_custom.yaml
@@ -51,7 +51,7 @@ model:
   class_path: eva.HeadModule
   init_args:
     head:
-      class_path: batch_linear.batch_linear
+      class_path: eva.core.models.networks.batch_linear.batch_linear
       init_args:
         in_features: ${oc.env:IN_FEATURES, 1280} #MODIFY ME
         out_features: &NUM_CLASSES 4
diff --git a/src/eva/core/data/datasets/classification/embeddings.py b/src/eva/core/data/datasets/classification/embeddings.py
index 489b20dd2..1ef4ce3de 100644
--- a/src/eva/core/data/datasets/classification/embeddings.py
+++ b/src/eva/core/data/datasets/classification/embeddings.py
@@ -26,7 +26,7 @@ def load_embeddings(self, index: int) -> torch.Tensor:
 
     @override
     def load_target(self, index: int) -> torch.Tensor:
-        target = self._data.at[index, self._column_mapping["target"]]
+        target = int(self._data.at[index, self._column_mapping["target"]])
         return torch.tensor(target, dtype=torch.int64)
 
     @override
diff --git a/src/eva/core/models/networks/adj_pooled_linear.py b/src/eva/core/models/networks/adj_pooled_linear.py
new file mode 100644
index 000000000..02fb422f2
--- /dev/null
+++ b/src/eva/core/models/networks/adj_pooled_linear.py
@@ -0,0 +1,61 @@
+import torch
+from torch import nn
+import math
+
+class adj_pooled_linear(nn.Module):
+
+    def __init__(
+        self,
+        in_features: int,
+        out_features: int,
+        pool_kernel_size = (7, 7),
+        patch_grid_hw = (14, 14),
+    ):
+        super().__init__()
+
+        self.in_features = in_features
+        self.grid_h, self.grid_w = patch_grid_hw
+        self.n_patches = self.grid_h * self.grid_w
+
+        if self.grid_h <= 0 or self.grid_w <= 0:
+            raise ValueError(f"patch_grid_hw must be positive, got {patch_grid_hw}.")
+
+        k_h, k_w = pool_kernel_size
+        if (self.grid_h % k_h) != 0 or (self.grid_w % k_w) != 0:
+            raise ValueError(f"pool_kernel_size {pool_kernel_size} must tile patch_grid_hw {patch_grid_hw}.")
+
+        self.pooled_h = self.grid_h // k_h
+        self.pooled_w = self.grid_w // k_w
+        pooled_n_patches = self.pooled_h * self.pooled_w
+
+        self.batch = nn.RMSNorm(in_features)
+        self.pool = nn.AvgPool2d(kernel_size=pool_kernel_size, stride=pool_kernel_size)
+
+        self.cls_lin = nn.Linear(in_features, out_features)
+        self.patch_linear = nn.Linear(in_features * pooled_n_patches, out_features)
+        self.reg_linear = nn.Linear(in_features * 4, out_features)
+
+    def forward(self, x):
+
+        x = self.batch(x)
+
+        B = x.shape[0]
+
+        registers = x[:, 0:4, :]
+        patch = x[:, 4:-1, :]
+        cls = x[:, -1, :]
+
+        patch_grid = patch.reshape(B, self.grid_h, self.grid_w, self.in_features).permute(0, 3, 1, 2)
+        pooled_patches_flat = self.pool(patch_grid).reshape(B, -1)
+
+        registers_flat = registers.reshape(B, -1)
+
+        registers_out = self.reg_linear(registers_flat)
+        patch_out = self.patch_linear(pooled_patches_flat)
+        cls_out = self.cls_lin(cls)
+
+        xs = torch.stack((registers_out, patch_out, cls_out), dim=1)
+
+        x_out = torch.mean(xs, dim=1)
+
+        return x_out
diff --git a/src/eva/core/models/networks/adj_pooled_linear_noreg.py b/src/eva/core/models/networks/adj_pooled_linear_noreg.py
new file mode 100644
index 000000000..d42932cb9
--- /dev/null
+++ b/src/eva/core/models/networks/adj_pooled_linear_noreg.py
@@ -0,0 +1,56 @@
+import torch
+from torch import nn
+import math
+
+class adj_pooled_linear_noreg(nn.Module):
+
+    def __init__(
+        self,
+        in_features: int,
+        out_features: int,
+        pool_kernel_size = (7, 7),
+        patch_grid_hw = (14, 14),
+    ):
+        super().__init__()
+
+        self.in_features = in_features
+        self.grid_h, self.grid_w = patch_grid_hw
+        self.n_patches = self.grid_h * self.grid_w
+
+        if self.grid_h <= 0 or self.grid_w <= 0:
+            raise ValueError(f"patch_grid_hw must be positive, got {patch_grid_hw}.")
+
+        k_h, k_w = pool_kernel_size
+        if (self.grid_h % k_h) != 0 or (self.grid_w % k_w) != 0:
+            raise ValueError(f"pool_kernel_size {pool_kernel_size} must tile patch_grid_hw {patch_grid_hw}.")
+
+        self.pooled_h = self.grid_h // k_h
+        self.pooled_w = self.grid_w // k_w
+        pooled_n_patches = self.pooled_h * self.pooled_w
+
+        self.batch = nn.RMSNorm(in_features)
+        self.pool = nn.AvgPool2d(kernel_size=pool_kernel_size, stride=pool_kernel_size)
+
+        self.cls_lin = nn.Linear(in_features, out_features)
+        self.patch_linear = nn.Linear(in_features * pooled_n_patches, out_features)
+
+    def forward(self, x):
+
+        x = self.batch(x)
+
+        B = x.shape[0]
+
+        patch = x[:, 4:-1, :]
+        cls = x[:, -1, :]
+
+        patch_grid = patch.reshape(B, self.grid_h, self.grid_w, self.in_features).permute(0, 3, 1, 2)
+        pooled_patches_flat = self.pool(patch_grid).reshape(B, -1)
+
+        patch_out = self.patch_linear(pooled_patches_flat)
+        cls_out = self.cls_lin(cls)
+
+        xs = torch.stack((patch_out, cls_out), dim=1)
+
+        x_out = torch.mean(xs, dim=1)
+
+        return x_out
diff --git a/src/eva/core/models/networks/batch_linear.py b/src/eva/core/models/networks/batch_linear.py
new file mode 100644
index 000000000..6490c76a2
--- /dev/null
+++ b/src/eva/core/models/networks/batch_linear.py
@@ -0,0 +1,23 @@
+
+
+from torch import nn
+
+class batch_linear(nn.Module):
+
+    def __init__(self, in_features: int, out_features: int):
+        super().__init__()
+
+        #self.batch = nn.BatchNorm1d(in_features)
+        self.batch = nn.RMSNorm(in_features)
+        self.lin = nn.Linear(in_features, out_features)
+
+
+    def forward(self, x):
+
+        x = self.batch(x)
+        x = self.lin(x)
+        x = x[:, -1, :]
+
+        return x
+
+
diff --git a/src/eva/core/models/networks/three_linear.py b/src/eva/core/models/networks/three_linear.py
new file mode 100644
index 000000000..d208ab887
--- /dev/null
+++ b/src/eva/core/models/networks/three_linear.py
@@ -0,0 +1,43 @@
+
+import torch
+from torch import nn
+
+class three_linear(nn.Module):
+
+    def __init__(self, in_features: int, out_features: int):
+        super().__init__()
+
+        #self.batch = nn.BatchNorm1d(out_features)
+        #self.batch = nn.RMSNorm(out_features)
+
+        self.cls_lin = nn.Linear(in_features, out_features)
+        self.patch_linear = nn.Linear(in_features * 196, out_features)
+        self.reg_linear = nn.Linear(in_features * 4, out_features)
+
+    def forward(self, x):
+        
+        #4 register, 196 patch, 1 cls
+        registers = x[:, 0:4, :]
+        registers = registers.reshape(x.shape[0], -1)#Flatten
+
+        patch = x[:, 4:-1, :]
+        patch = patch.reshape(x.shape[0], -1)
+
+        cls = x[:, -1 , :]
+
+
+        registers = self.reg_linear(registers)
+        patch = self.patch_linear(patch)
+        cls = self.cls_lin(cls)
+        
+        
+        xs = torch.stack((registers, patch, cls), axis = 1)
+        
+        #output shape is batch, 3, embedd
+        
+        #Either average the features, or learn some kind of combiner.
+        x = torch.mean(xs, dim = 1)
+#        x = self.batch(x)
+
+        return x
+
diff --git a/src/eva/core/models/networks/tied_double_linear.py b/src/eva/core/models/networks/tied_double_linear.py
new file mode 100644
index 000000000..09f62ac87
--- /dev/null
+++ b/src/eva/core/models/networks/tied_double_linear.py
@@ -0,0 +1,46 @@
+import torch
+from torch import nn
+
+class tied_double_linear(nn.Module):
+
+    def __init__(
+        self,
+        in_features: int,
+        out_features: int,
+        projection_dim: int = 64,
+        n_patches: int = 196
+    ):
+
+        super().__init__()
+
+        self.batch = nn.RMSNorm(in_features)
+        self.projection = nn.Linear(in_features, projection_dim)
+
+        self.cls_lin = nn.Linear(in_features, out_features)
+        self.patch_linear = nn.Linear(projection_dim * n_patches, out_features)
+        self.reg_linear = nn.Linear(in_features * 4, out_features)
+
+    def forward(self, x):
+
+        x = self.batch(x)
+
+        B = x.shape[0]
+
+        registers = x[:, 0:4, :]
+        patch = x[:, 4:-1, :]
+        cls = x[:, -1, :]
+
+        projected_patches = self.projection(patch)
+        projected_patches_flat = projected_patches.reshape(B, -1)
+
+        registers_flat = registers.reshape(B, -1)
+
+        registers_out = self.reg_linear(registers_flat)
+        patch_out = self.patch_linear(projected_patches_flat)
+        cls_out = self.cls_lin(cls)
+
+        xs = torch.stack((registers_out, patch_out, cls_out), dim=1)
+
+        x_out = torch.mean(xs, dim=1)
+
+        return x_out