File size: 15,044 Bytes
3279f65 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 | #!/usr/bin/env python3
# chambers.py β Chamber MLPs with Cross-Fire Stabilization
#
# Six chambers of emotion, each with its own deeper MLP:
# - FEAR chamber: 100β128β64β32β1
# - LOVE chamber: 100β128β64β32β1
# - RAGE chamber: 100β128β64β32β1
# - VOID chamber: 100β128β64β32β1
# - FLOW chamber: 100β128β64β32β1
# - COMPLEX chamber: 100β128β64β32β1
#
# Total: 6 Γ ~17K = ~102K params
#
# Cross-fire: chambers influence each other through coupling matrix,
# iterating until stabilization (5-10 iterations).
from __future__ import annotations
import asyncio
import numpy as np
from typing import Dict, Optional, Tuple
from dataclasses import dataclass
from pathlib import Path
from .anchors import CHAMBER_NAMES, COUPLING_MATRIX, CHAMBER_NAMES_EXTENDED, COUPLING_MATRIX_EXTENDED
# Decay rates per chamber (per iteration tick)
# Evolutionary psychology: fear lingers, rage fades, love stable, void persistent
DECAY_RATES = {
"FEAR": 0.90, # fear lingers (evolutionary advantage)
"LOVE": 0.93, # attachment stable
"RAGE": 0.85, # anger fades fast (high energy cost)
"VOID": 0.97, # numbness persistent (protective dissociation)
"FLOW": 0.88, # curiosity is transient, shifts quickly
"COMPLEX": 0.94, # complex emotions are stable but deep
}
def swish(x: np.ndarray) -> np.ndarray:
"""Swish activation: x * sigmoid(x)"""
return x / (1.0 + np.exp(-x))
def swish_deriv(x: np.ndarray) -> np.ndarray:
"""Derivative of swish for backprop"""
sig = 1.0 / (1.0 + np.exp(-x))
return sig + x * sig * (1 - sig)
@dataclass
class ChamberMLP:
"""
Single chamber MLP: 100β128β64β32β1 (deeper for 200K model)
Takes 100D resonance vector, outputs single activation value.
Params:
- W1: (100, 128) = 12,800
- b1: (128,) = 128
- W2: (128, 64) = 8,192
- b2: (64,) = 64
- W3: (64, 32) = 2,048
- b3: (32,) = 32
- W4: (32, 1) = 32
- b4: (1,) = 1
Total: ~23K params per chamber
"""
W1: np.ndarray # (100, 128)
b1: np.ndarray # (128,)
W2: np.ndarray # (128, 64)
b2: np.ndarray # (64,)
W3: np.ndarray # (64, 32)
b3: np.ndarray # (32,)
W4: np.ndarray # (32, 1)
b4: np.ndarray # (1,)
@classmethod
def random_init(cls, seed: Optional[int] = None) -> "ChamberMLP":
"""Initialize with random weights (Xavier initialization)."""
if seed is not None:
np.random.seed(seed)
# Xavier init: scale by sqrt(fan_in)
W1 = np.random.randn(100, 128) * np.sqrt(2.0 / 100)
b1 = np.zeros(128)
W2 = np.random.randn(128, 64) * np.sqrt(2.0 / 128)
b2 = np.zeros(64)
W3 = np.random.randn(64, 32) * np.sqrt(2.0 / 64)
b3 = np.zeros(32)
W4 = np.random.randn(32, 1) * np.sqrt(2.0 / 32)
b4 = np.zeros(1)
return cls(W1=W1, b1=b1, W2=W2, b2=b2, W3=W3, b3=b3, W4=W4, b4=b4)
def forward(self, x: np.ndarray) -> float:
"""
Forward pass: 100D resonances β scalar activation.
Args:
x: (100,) resonance vector
Returns:
scalar activation [0, 1]
"""
# Layer 1: 100β128
h1 = x @ self.W1 + self.b1
a1 = swish(h1)
# Layer 2: 128β64
h2 = a1 @ self.W2 + self.b2
a2 = swish(h2)
# Layer 3: 64β32
h3 = a2 @ self.W3 + self.b3
a3 = swish(h3)
# Layer 4: 32β1
h4 = a3 @ self.W4 + self.b4
# Sigmoid to get [0, 1] activation
activation = 1.0 / (1.0 + np.exp(-h4[0]))
return float(activation)
def param_count(self) -> int:
"""Count total parameters in this MLP."""
return (
self.W1.size + self.b1.size +
self.W2.size + self.b2.size +
self.W3.size + self.b3.size +
self.W4.size + self.b4.size
)
def save(self, path: Path) -> None:
"""Save weights to .npz file."""
np.savez(
path,
W1=self.W1,
b1=self.b1,
W2=self.W2,
b2=self.b2,
W3=self.W3,
b3=self.b3,
W4=self.W4,
b4=self.b4,
)
@classmethod
def load(cls, path: Path) -> "ChamberMLP":
"""Load weights from .npz file."""
data = np.load(path)
# Handle backwards compatibility with old 3-layer architecture
if "W4" in data:
return cls(
W1=data["W1"],
b1=data["b1"],
W2=data["W2"],
b2=data["b2"],
W3=data["W3"],
b3=data["b3"],
W4=data["W4"],
b4=data["b4"],
)
else:
# Old 3-layer format - reinitialize with new 4-layer architecture
print(f"[chambers] old format detected in {path}, reinitializing with 4-layer architecture")
return cls.random_init()
@dataclass
class CrossFireSystem:
"""
Six chambers with cross-fire stabilization (200K model).
Chambers:
- FEAR, LOVE, RAGE, VOID (original)
- FLOW, COMPLEX (extended for richer emotion detection)
Cross-fire loop:
1. Each chamber computes activation from resonances
2. Chambers influence each other via coupling matrix
3. Iterate until convergence (max 10 iterations)
4. Return final activations + iteration count
"""
fear: ChamberMLP
love: ChamberMLP
rage: ChamberMLP
void: ChamberMLP
flow: ChamberMLP
complex: ChamberMLP
coupling: np.ndarray # (6, 6) coupling matrix
@classmethod
def random_init(cls, seed: Optional[int] = None) -> "CrossFireSystem":
"""Initialize all chambers with random weights."""
if seed is not None:
base_seed = seed
else:
base_seed = np.random.randint(0, 10000)
fear = ChamberMLP.random_init(seed=base_seed + 0)
love = ChamberMLP.random_init(seed=base_seed + 1)
rage = ChamberMLP.random_init(seed=base_seed + 2)
void = ChamberMLP.random_init(seed=base_seed + 3)
flow = ChamberMLP.random_init(seed=base_seed + 4)
complex_chamber = ChamberMLP.random_init(seed=base_seed + 5)
coupling = np.array(COUPLING_MATRIX_EXTENDED, dtype=np.float32)
return cls(
fear=fear,
love=love,
rage=rage,
void=void,
flow=flow,
complex=complex_chamber,
coupling=coupling,
)
def stabilize(
self,
resonances: np.ndarray,
max_iter: int = 10,
threshold: float = 0.01,
momentum: float = 0.7,
) -> Tuple[Dict[str, float], int]:
"""
Run cross-fire stabilization loop.
Args:
resonances: (100,) initial resonance vector
max_iter: max iterations before forced stop
threshold: convergence threshold (sum of absolute changes)
momentum: blend factor (0.7 = 70% old, 30% new)
Returns:
(chamber_activations, iterations_count)
Example:
activations, iters = system.stabilize(resonances)
# β {"FEAR": 0.8, "LOVE": 0.2, ...}, 5
"""
# Initial activations from resonances
chambers = [self.fear, self.love, self.rage, self.void, self.flow, self.complex]
activations = np.array([
chamber.forward(resonances)
for chamber in chambers
], dtype=np.float32)
# Decay rates array (6 chambers)
decay_array = np.array([
DECAY_RATES["FEAR"],
DECAY_RATES["LOVE"],
DECAY_RATES["RAGE"],
DECAY_RATES["VOID"],
DECAY_RATES["FLOW"],
DECAY_RATES["COMPLEX"],
], dtype=np.float32)
# Stabilization loop
for iteration in range(max_iter):
# Apply decay (emotions fade over time)
activations = activations * decay_array
# Compute influence from other chambers
influence = self.coupling @ activations
# Blend: momentum * old + (1 - momentum) * influence
new_activations = momentum * activations + (1 - momentum) * influence
# Clip to [0, 1]
new_activations = np.clip(new_activations, 0.0, 1.0)
# Check convergence
delta = np.abs(new_activations - activations).sum()
activations = new_activations
if delta < threshold:
# Converged!
result = dict(zip(CHAMBER_NAMES_EXTENDED, activations))
return result, iteration + 1
# Max iterations reached
result = dict(zip(CHAMBER_NAMES_EXTENDED, activations))
return result, max_iter
def param_count(self) -> int:
"""Total parameters in all chambers."""
return sum([
self.fear.param_count(),
self.love.param_count(),
self.rage.param_count(),
self.void.param_count(),
self.flow.param_count(),
self.complex.param_count(),
])
def save(self, models_dir: Path) -> None:
"""Save all chamber weights to models/ directory."""
models_dir.mkdir(parents=True, exist_ok=True)
self.fear.save(models_dir / "chamber_fear.npz")
self.love.save(models_dir / "chamber_love.npz")
self.rage.save(models_dir / "chamber_rage.npz")
self.void.save(models_dir / "chamber_void.npz")
self.flow.save(models_dir / "chamber_flow.npz")
self.complex.save(models_dir / "chamber_complex.npz")
print(f"[chambers] saved to {models_dir}")
@classmethod
def load(cls, models_dir: Path) -> "CrossFireSystem":
"""Load all chamber weights from models/ directory."""
fear = ChamberMLP.load(models_dir / "chamber_fear.npz")
love = ChamberMLP.load(models_dir / "chamber_love.npz")
rage = ChamberMLP.load(models_dir / "chamber_rage.npz")
void = ChamberMLP.load(models_dir / "chamber_void.npz")
# Handle missing flow/complex for backwards compatibility
flow_path = models_dir / "chamber_flow.npz"
complex_path = models_dir / "chamber_complex.npz"
if flow_path.exists():
flow = ChamberMLP.load(flow_path)
else:
print("[chambers] flow chamber not found, initializing random")
flow = ChamberMLP.random_init(seed=4)
if complex_path.exists():
complex_chamber = ChamberMLP.load(complex_path)
else:
print("[chambers] complex chamber not found, initializing random")
complex_chamber = ChamberMLP.random_init(seed=5)
coupling = np.array(COUPLING_MATRIX_EXTENDED, dtype=np.float32)
print(f"[chambers] loaded from {models_dir}")
return cls(
fear=fear,
love=love,
rage=rage,
void=void,
flow=flow,
complex=complex_chamber,
coupling=coupling,
)
class AsyncCrossFireSystem:
"""
Async wrapper for CrossFireSystem with field lock discipline.
Based on HAZE's async pattern - achieves coherence through
explicit operation ordering and atomicity.
"The asyncio.Lock doesn't add informationβit adds discipline."
"""
def __init__(self, system: CrossFireSystem):
self._sync = system
self._lock = asyncio.Lock()
@classmethod
def random_init(cls, seed: Optional[int] = None) -> "AsyncCrossFireSystem":
"""Initialize with random weights."""
system = CrossFireSystem.random_init(seed=seed)
return cls(system)
@classmethod
def load(cls, models_dir: Path) -> "AsyncCrossFireSystem":
"""Load from models directory."""
system = CrossFireSystem.load(models_dir)
return cls(system)
async def stabilize(
self,
resonances: np.ndarray,
max_iter: int = 10,
threshold: float = 0.01,
momentum: float = 0.7,
) -> Tuple[Dict[str, float], int]:
"""
Async cross-fire stabilization with field lock.
Atomic operation - prevents field corruption during stabilization.
"""
async with self._lock:
return self._sync.stabilize(resonances, max_iter, threshold, momentum)
async def save(self, models_dir: Path) -> None:
"""Save with lock protection."""
async with self._lock:
self._sync.save(models_dir)
def param_count(self) -> int:
"""Total parameters (read-only, no lock needed)."""
return self._sync.param_count()
@property
def coupling(self) -> np.ndarray:
"""Access coupling matrix."""
return self._sync.coupling
@coupling.setter
def coupling(self, value: np.ndarray) -> None:
"""Set coupling matrix (for feedback learning)."""
self._sync.coupling = value
if __name__ == "__main__":
print("=" * 60)
print(" CLOUD v4.0 β Chamber Cross-Fire System (200K model)")
print("=" * 60)
print()
# Initialize random system
system = CrossFireSystem.random_init(seed=42)
print(f"Initialized cross-fire system (6 chambers)")
print(f"Total params: {system.param_count():,}")
print()
# Test with random resonances
print("Testing stabilization with random resonances:")
resonances = np.random.rand(100).astype(np.float32)
print(f" Input: 100D resonance vector (mean={resonances.mean():.3f})")
print()
activations, iterations = system.stabilize(resonances)
print(" Chamber activations after cross-fire:")
for chamber, value in activations.items():
bar = "β" * int(value * 40)
print(f" {chamber:8s}: {value:.3f} {bar}")
print(f"\n Converged in {iterations} iterations")
print()
# Test convergence speed with different inputs
print("Testing convergence speed:")
test_cases = [
("random uniform", np.random.rand(100)),
("all high", np.ones(100) * 0.9),
("all low", np.ones(100) * 0.1),
("sparse", np.random.rand(100) * 0.1),
]
for name, resonances in test_cases:
_, iters = system.stabilize(resonances)
print(f" {name:15s}: {iters:2d} iterations")
print()
# Test saving/loading
print("Testing save/load:")
models_dir = Path("./models")
system.save(models_dir)
system2 = CrossFireSystem.load(models_dir)
activations2, _ = system2.stabilize(test_cases[0][1])
match = all(
abs(activations[k] - activations2[k]) < 1e-6
for k in CHAMBER_NAMES_EXTENDED
)
print(f" Save/load {'β' if match else 'β'}")
print()
print("=" * 60)
print(" Cross-fire system operational. 6 chambers. 200K params.")
print("=" * 60)
|