Upload model

config.json

@@ -0,0 +1,20 @@
+{
+  "architectures": [
+    "SpiceCNNModelForImageClassification"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_spice_cnn.SpiceCNNConfig",
+    "AutoModelForImageClassification": "modeling_spice_cnn.SpiceCNNModelForImageClassification"
+  },
+  "dropout_rate": 0.2,
+  "hidden_size": 128,
+  "kernel_size": 3,
+  "model_type": "spicecnn",
+  "num_classes": 10,
+  "num_filters": 16,
+  "padding": 1,
+  "pooling_size": 2,
+  "stride": 1,
+  "torch_dtype": "float32",
+  "transformers_version": "4.29.2"
+}

configuration_spice_cnn.py

@@ -0,0 +1,48 @@
+from transformers import PretrainedConfig
+
+"""Spice CNN model configuration"""
+
+SPICE_CNN_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "spicecloud/spice-cnn-base": "https://huggingface.co/spice-cnn-base/resolve/main/config.json"
+}
+
+
+# Define custom convnet configuration
+class SpiceCNNConfig(PretrainedConfig):
+    """
+    This is the configuration class to store the configuration of a [`SpiceCNNModel`].
+    It is used to instantiate an SpiceCNN model according to the specified arguments,
+    defining the model architecture. Instantiating a configuration with the defaults
+    will yield a similar configuration to that of the SpiceCNN
+    [spicecloud/spice-cnn-base](https://huggingface.co/spicecloud/spice-cnn-base)
+    architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control
+    the model outputs. Read the documentation from [`PretrainedConfig`] for more
+    information.
+    """
+
+    model_type = "spicecnn"
+
+    def __init__(
+        self,
+        num_classes: int = 10,
+        dropout_rate: float = 0.2,
+        hidden_size: int = 128,
+        num_filters: int = 16,
+        kernel_size: int = 3,
+        stride: int = 1,
+        padding: int = 1,
+        pooling_size: int = 2,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+
+        self.num_classes = num_classes
+        self.dropout_rate = dropout_rate
+        self.hidden_size = hidden_size
+        self.num_filters = num_filters
+        self.kernel_size = kernel_size
+        self.stride = stride
+        self.padding = padding
+        self.pooling_size = pooling_size

modeling_spice_cnn.py

@@ -0,0 +1,44 @@
+import torch.nn as nn
+
+from transformers import PreTrainedModel
+
+from .configuration_spice_cnn import SpiceCNNConfig
+
+
+class SpiceCNNModelForImageClassification(PreTrainedModel):
+    config_class = SpiceCNNConfig
+
+    def __init__(self, config: SpiceCNNConfig):
+        super().__init__(config)
+        layers = [
+            nn.Conv2d(
+                1,
+                32,
+                kernel_size=config.kernel_size,
+                stride=config.stride,
+                padding=config.padding,
+            ),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=config.pooling_size),
+            nn.Conv2d(
+                32,
+                64,
+                kernel_size=config.kernel_size,
+                stride=config.stride,
+                padding=config.padding,
+            ),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=config.pooling_size),
+            nn.Flatten(),
+            nn.Linear(7 * 7 * 64, 128),
+            nn.ReLU(),
+            nn.Linear(128, config.num_classes),
+        ]
+        self.model = nn.Sequential(*layers)
+
+    def forward(self, tensor, labels=None):
+        logits = self.model(tensor)
+        if labels is not None:
+            loss = nn.CrossEntropyLoss(logits, labels)
+            return {"loss": loss, "logits": logits}
+        return {"logits": logits}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb3719d131ffeda1fe4d66a097fad200329816c955394cadbecc1f849f7438e7
+size 1689227