JK flip-flop

README.md

@@ -0,0 +1,92 @@
+---
+license: mit
+tags:
+- pytorch
+- safetensors
+- threshold-logic
+- neuromorphic
+- sequential
+- flipflop
+---
+
+# threshold-jk-flipflop
+
+JK flip-flop next-state logic as threshold circuit. Most versatile flip-flop with Set, Reset, Hold, and Toggle modes.
+
+## Circuit
+
+```
+J      ──┐
+K      ──┼──► JK-FF ──┬──► Q
+Q_prev ──┘           └──► Qn
+```
+
+## Truth Table
+
+| J | K | Q_prev | Q | Qn | Mode |
+|---|---|--------|---|----|----|
+| 0 | 0 | 0 | 0 | 1 | Hold |
+| 0 | 0 | 1 | 1 | 0 | Hold |
+| 0 | 1 | X | 0 | 1 | Reset |
+| 1 | 0 | X | 1 | 0 | Set |
+| 1 | 1 | 0 | 1 | 0 | Toggle |
+| 1 | 1 | 1 | 0 | 1 | Toggle |
+
+## Modes
+
+```
+J=0, K=0: Hold    - Q stays same
+J=0, K=1: Reset   - Q becomes 0
+J=1, K=0: Set     - Q becomes 1
+J=1, K=1: Toggle  - Q flips
+```
+
+## Logic
+
+```
+Q  = J·~Q_prev + ~K·Q_prev
+Qn = ~J·~Q_prev + K·Q_prev
+```
+
+## Architecture
+
+| Component | Neurons |
+|-----------|---------|
+| Input inversions | 4 |
+| AND gates | 4 |
+| OR outputs | 2 |
+| Passthroughs | 2 |
+
+**Total: 12 neurons, 42 parameters, 3 layers**
+
+## Advantage Over SR
+
+Unlike SR flip-flop, JK has no invalid state. When both inputs are high, it toggles instead of producing undefined behavior.
+
+## Usage
+
+```python
+from safetensors.torch import load_file
+
+w = load_file('model.safetensors')
+
+# Counter using toggle mode:
+q = 0
+for _ in range(4):
+    q_next = compute(j=1, k=1, q_prev=q, w)
+    q = q_next  # q toggles: 0,1,0,1,...
+```
+
+## Files
+
+```
+threshold-jk-flipflop/
+├── model.safetensors
+├── create_safetensors.py
+├── config.json
+└── README.md
+```
+
+## License
+
+MIT

config.json

@@ -0,0 +1,9 @@
+{
+  "name": "threshold-jk-flipflop",
+  "description": "JK flip-flop next-state logic as threshold circuit",
+  "inputs": 3,
+  "outputs": 2,
+  "neurons": 12,
+  "layers": 3,
+  "parameters": 42
+}

create_safetensors.py

@@ -0,0 +1,127 @@
+import torch
+from safetensors.torch import save_file
+
+weights = {}
+
+# JK Flip-Flop
+# Inputs: J, K, Q_prev (3 inputs)
+# Outputs: Q, Qn (2 outputs)
+#
+# J=0, K=0: Hold (Q = Q_prev)
+# J=0, K=1: Reset (Q = 0)
+# J=1, K=0: Set (Q = 1)
+# J=1, K=1: Toggle (Q = ~Q_prev)
+#
+# Q = J·~Q_prev + ~K·Q_prev
+
+# Input indices: J=0, K=1, Q_prev=2
+
+# NOT Q_prev
+weights['not_qprev.weight'] = torch.tensor([[0.0, 0.0, -1.0]], dtype=torch.float32)
+weights['not_qprev.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# J passthrough
+weights['j.weight'] = torch.tensor([[1.0, 0.0, 0.0]], dtype=torch.float32)
+weights['j.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# NOT K
+weights['not_k.weight'] = torch.tensor([[0.0, -1.0, 0.0]], dtype=torch.float32)
+weights['not_k.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# Q_prev passthrough
+weights['qprev.weight'] = torch.tensor([[0.0, 0.0, 1.0]], dtype=torch.float32)
+weights['qprev.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# J AND NOT Q_prev
+weights['j_and_notq.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['j_and_notq.bias'] = torch.tensor([-2.0], dtype=torch.float32)
+
+# NOT K AND Q_prev
+weights['notk_and_q.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['notk_and_q.bias'] = torch.tensor([-2.0], dtype=torch.float32)
+
+# Q = (J AND NOT Q_prev) OR (NOT K AND Q_prev)
+weights['q.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['q.bias'] = torch.tensor([-1.0], dtype=torch.float32)
+
+# Qn = ~J·~Q_prev + K·Q_prev
+# NOT J
+weights['not_j.weight'] = torch.tensor([[-1.0, 0.0, 0.0]], dtype=torch.float32)
+weights['not_j.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# K passthrough
+weights['k.weight'] = torch.tensor([[0.0, 1.0, 0.0]], dtype=torch.float32)
+weights['k.bias'] = torch.tensor([0.0], dtype=torch.float32)
+
+# NOT J AND NOT Q_prev
+weights['notj_and_notq.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['notj_and_notq.bias'] = torch.tensor([-2.0], dtype=torch.float32)
+
+# K AND Q_prev
+weights['k_and_q.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['k_and_q.bias'] = torch.tensor([-2.0], dtype=torch.float32)
+
+# Qn = (NOT J AND NOT Q_prev) OR (K AND Q_prev)
+weights['qn.weight'] = torch.tensor([[1.0, 1.0]], dtype=torch.float32)
+weights['qn.bias'] = torch.tensor([-1.0], dtype=torch.float32)
+
+save_file(weights, 'model.safetensors')
+
+def jk_ff(j, k, q_prev):
+    if j == 0 and k == 0:
+        return q_prev, 1 - q_prev
+    elif j == 0 and k == 1:
+        return 0, 1
+    elif j == 1 and k == 0:
+        return 1, 0
+    else:
+        return 1 - q_prev, q_prev
+
+def eval_circuit(j, k, q_prev):
+    not_qprev = 1 - q_prev
+    not_k = 1 - k
+    j_and_notq = j and not_qprev
+    notk_and_q = not_k and q_prev
+    q = j_and_notq or notk_and_q
+    not_j = 1 - j
+    notj_and_notq = not_j and not_qprev
+    k_and_q = k and q_prev
+    qn = notj_and_notq or k_and_q
+    return q, qn
+
+print("Verifying JK Flip-Flop...")
+errors = 0
+for j in range(2):
+    for k in range(2):
+        for q_prev in range(2):
+            expected_q, expected_qn = jk_ff(j, k, q_prev)
+            calc_q, calc_qn = eval_circuit(j, k, q_prev)
+            if expected_q != calc_q or expected_qn != calc_qn:
+                errors += 1
+                print(f"ERROR: J={j}, K={k}, Q_prev={q_prev}")
+
+if errors == 0:
+    print("All 8 test cases passed!")
+else:
+    print(f"FAILED: {errors} errors")
+
+print("\nTruth Table:")
+print("J K Q_prev | Q Qn | Mode")
+print("-" * 30)
+for j in range(2):
+    for k in range(2):
+        for q_prev in range(2):
+            q, qn = jk_ff(j, k, q_prev)
+            if j == 0 and k == 0:
+                mode = "Hold"
+            elif j == 0 and k == 1:
+                mode = "Reset"
+            elif j == 1 and k == 0:
+                mode = "Set"
+            else:
+                mode = "Toggle"
+            print(f"{j} {k}   {q_prev}    | {q}  {qn} | {mode}")
+
+mag = sum(t.abs().sum().item() for t in weights.values())
+print(f"\nMagnitude: {mag:.0f}")
+print(f"Parameters: {sum(t.numel() for t in weights.values())}")