Upload folder using huggingface_hub

README.md

@@ -0,0 +1,63 @@
+---
+license: mit
+tags:
+- pytorch
+- safetensors
+- threshold-logic
+- neuromorphic
+---
+
+# threshold-nand4
+
+4-input NAND gate. Outputs 0 only when all inputs are 1.
+
+## Circuit
+
+```
+  x1  x2  x3  x4
+   │   │   │   │
+   └───┴───┴───┘
+         │
+         ▼
+    ┌──────────┐
+    │w:-1,-1,-1,-1│
+    │ b: 3     │
+    └──────────┘
+         │
+         ▼
+  NAND4(x1,x2,x3,x4)
+```
+
+## Parameters
+
+| | |
+|---|---|
+| Weights | [-1, -1, -1, -1] |
+| Bias | 3 |
+| Magnitude | 7 |
+
+## Optimality
+
+Exhaustive enumeration of 7,183 configurations confirms magnitude 7 is optimal. 1 valid configuration exists.
+
+| Magnitude | Valid Configs |
+|-----------|---------------|
+| 0-6 | 0 |
+| 7 | 1 |
+
+## Usage
+
+```python
+from safetensors.torch import load_file
+import torch
+
+w = load_file('model.safetensors')
+
+def nand4(x1, x2, x3, x4):
+    inputs = torch.tensor([float(x1), float(x2), float(x3), float(x4)])
+    return int((inputs * w['weight']).sum() + w['bias'] >= 0)
+```
+
+## License
+
+MIT

config.json

@@ -0,0 +1,19 @@
+{
+  "model_type": "threshold_network",
+  "task": "nand4_gate",
+  "architecture": "4 -> 1",
+  "input_size": 4,
+  "output_size": 1,
+  "num_neurons": 1,
+  "num_parameters": 5,
+  "activation": "heaviside",
+  "weight_constraints": "integer",
+  "verification": {
+    "method": "exhaustive",
+    "inputs_tested": 16
+  },
+  "accuracy": {
+    "all_inputs": "16/16",
+    "percentage": 100.0
+  }
+}

create_safetensors.py

@@ -0,0 +1,13 @@
+import torch
+from safetensors.torch import save_file
+
+weights = {
+    'weight': torch.tensor([-1.0, -1.0, -1.0, -1.0]),
+    'bias': torch.tensor([3.0]),
+}
+
+save_file(weights, 'model.safetensors')
+print('Created model.safetensors')
+print(f'  weight: {weights["weight"].tolist()}')
+print(f'  bias: {weights["bias"].item()}')
+print(f'  magnitude: {sum(abs(w) for w in weights["weight"].tolist()) + abs(weights["bias"].item()):.0f}')

model.py

@@ -0,0 +1,41 @@
+"""
+Threshold Network for 4-input NAND Gate
+"""
+
+import torch
+from safetensors.torch import load_file
+
+
+class ThresholdNAND4:
+    def __init__(self, weights_dict):
+        self.weight = weights_dict['weight']
+        self.bias = weights_dict['bias']
+
+    def __call__(self, x1, x2, x3, x4):
+        inputs = torch.tensor([float(x1), float(x2), float(x3), float(x4)])
+        weighted_sum = (inputs * self.weight).sum() + self.bias
+        return (weighted_sum >= 0).float()
+
+    @classmethod
+    def from_safetensors(cls, path="model.safetensors"):
+        return cls(load_file(path))
+
+
+if __name__ == "__main__":
+    weights = load_file("model.safetensors")
+    model = ThresholdNAND4(weights)
+
+    print("4-input NAND Gate Truth Table:")
+    print("-" * 35)
+    correct = 0
+    for x1 in [0, 1]:
+        for x2 in [0, 1]:
+            for x3 in [0, 1]:
+                for x4 in [0, 1]:
+                    out = int(model(x1, x2, x3, x4).item())
+                    expected = 1 - (x1 & x2 & x3 & x4)
+                    status = "OK" if out == expected else "FAIL"
+                    if out == expected:
+                        correct += 1
+                    print(f"NAND4({x1}, {x2}, {x3}, {x4}) = {out}  [{status}]")
+    print(f"\nTotal: {correct}/16 correct")

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55e8950d44e8e560b3aaa57e992904867a38321cd2a198a3d19934cedc796dd0
+size 148