File size: 12,255 Bytes
997a0f9 | 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 | """
AETHER Core: Central orchestrator integrating all subsystems.
Design: Neuro-Symbolic Fluidity + Constrained Self-Modification
"""
import torch
import torch.nn as nn
from typing import Dict, List, Any, Optional, Callable
import logging
from dataclasses import dataclass, field
import json
import hashlib
import time
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("AETHER.Core")
@dataclass
class AetherConfig:
"""Configuration for AETHER system evolution."""
# Evolution
population_size: int = 8
generations: int = 10
mutation_rate: float = 0.15
crossover_rate: float = 0.3
# Safety
sandbox_timeout: float = 30.0
max_architecture_depth: int = 5
require_human_approval: bool = False
# Hierarchical Reasoning
macro_policy_dim: int = 256
micro_policy_dim: int = 128
num_agents: int = 4
# Memory
working_memory_capacity: int = 16
episodic_buffer_size: int = 1000
# Knowledge
kg_embedding_dim: int = 128
kg_num_relations: int = 20
# Training
learning_rate: float = 2e-5
batch_size: int = 4
gradient_accumulation_steps: int = 8
# Meta
enable_self_modification: bool = True
enable_parallel_agents: bool = True
log_level: str = "INFO"
class AetherCore(nn.Module):
"""
Central controller for AETHER.
Manages the recursive evolution loop, agent orchestration,
knowledge integration, and safety constraints.
"""
def __init__(self, config: Optional[AetherConfig] = None,
model_name: str = "Qwen/Qwen2.5-0.5B-Instruct"):
super().__init__()
self.config = config or AetherConfig()
self.model_name = model_name
# Subsystems (initialized lazily)
self._memory = None
self._evolution = None
self._agents = None
self._knowledge = None
self._safety = None
# State tracking
self.generation = 0
self.architecture_history: List[Dict] = []
self.fitness_log: List[float] = []
self.metadata: Dict[str, Any] = {
"birth_timestamp": time.time(),
"model_name": model_name,
"version": "0.1.0",
}
# Neuro-symbolic bridge: learned attention over symbolic rules
self.symbolic_gate = nn.Parameter(torch.randn(1))
self.neural_gate = nn.Parameter(torch.randn(1))
logger.info(f"AETHER Core initialized with model: {model_name}")
@property
def memory(self):
if self._memory is None:
from .memory import CoALAMemory, TemporalMemory
self._memory = {
"working": CoALAMemory(capacity=self.config.working_memory_capacity),
"temporal": TemporalMemory(buffer_size=self.config.episodic_buffer_size),
}
return self._memory
@property
def evolution(self):
if self._evolution is None:
from .evolution import AetherEvolutionEngine
self._evolution = AetherEvolutionEngine(self.config)
return self._evolution
@property
def agents(self):
if self._agents is None:
from .agents import AetherAgentOrchestrator
self._agents = AetherAgentOrchestrator(self.config)
return self._agents
@property
def knowledge(self):
if self._knowledge is None:
from .knowledge import KnowledgeGraphEngine
self._knowledge = KnowledgeGraphEngine(
embedding_dim=self.config.kg_embedding_dim,
num_relations=self.config.kg_num_relations,
)
return self._knowledge
@property
def safety(self):
if self._safety is None:
from .safety import SafetySandbox
self._safety = SafetySandbox(timeout=self.config.sandbox_timeout)
return self._safety
def forward(self, task: str, context: Optional[Dict] = None) -> Dict[str, Any]:
"""
Main forward pass: given a task, orchestrate agents, query knowledge,
and produce output through neuro-symbolic fusion.
"""
context = context or {}
# 1. Retrieve relevant knowledge
kg_context = self.knowledge.query(task, top_k=5)
# 2. Load into working memory
self.memory["working"].store({
"task": task,
"kg_context": kg_context,
"timestamp": time.time(),
})
# 3. Hierarchical agent execution
result = self.agents.execute(task, kg_context, context)
# 4. Neuro-symbolic fusion gate
symbolic_weight = torch.sigmoid(self.symbolic_gate)
neural_weight = torch.sigmoid(self.neural_gate)
# Normalize
total = symbolic_weight + neural_weight + 1e-8
symbolic_weight = symbolic_weight / total
neural_weight = neural_weight / total
# 5. Store to episodic memory
self.memory["temporal"].store({
"task": task,
"result": result,
"weights": {
"symbolic": symbolic_weight.item(),
"neural": neural_weight.item(),
}
})
return {
"output": result,
"symbolic_weight": symbolic_weight.item(),
"neural_weight": neural_weight.item(),
"kg_context": kg_context,
"generation": self.generation,
}
def evolve(self, evaluation_function: Callable[[Dict], float],
num_generations: Optional[int] = None) -> Dict[str, Any]:
"""
Recursive evolutionary loop: generate candidates, evaluate,
select, mutate, validate, integrate.
Based on AlphaEvolve + GEA + ASI-Evolve methodology.
"""
num_generations = num_generations or self.config.generations
logger.info(f"Starting evolution for {num_generations} generations")
best_fitness = -float('inf')
best_config = None
for gen in range(num_generations):
self.generation = gen
# Generate candidate variants
candidates = self.evolution.generate_candidates(
base_config=self.config,
population_size=self.config.population_size,
)
# Evaluate each candidate
fitness_scores = []
for candidate in candidates:
# Safety sandbox evaluation
with self.safety.sandbox():
try:
score = evaluation_function(candidate)
fitness_scores.append(score)
except Exception as e:
logger.warning(f"Candidate failed evaluation: {e}")
fitness_scores.append(-float('inf'))
# Select top performers (Performance-Novelty from GEA)
selected = self.evolution.select(
candidates, fitness_scores,
alpha_exploration=0.3,
)
# Apply constrained mutations
mutated = self.evolution.mutate(
selected,
mutation_rate=self.config.mutation_rate,
max_depth=self.config.max_architecture_depth,
)
# Validate stability
validated = []
for candidate in mutated:
if self.safety.validate_architecture(candidate):
validated.append(candidate)
# Integrate best
if validated:
best_idx = max(range(len(validated)),
key=lambda i: fitness_scores[min(i, len(fitness_scores)-1)])
best_candidate = validated[best_idx]
current_fitness = fitness_scores[min(best_idx, len(fitness_scores)-1)]
if current_fitness > best_fitness:
best_fitness = current_fitness
best_config = best_candidate
self.config = best_candidate
# Log architecture change
arch_hash = hashlib.sha256(
json.dumps(best_candidate.__dict__, sort_keys=True).encode()
).hexdigest()[:16]
self.architecture_history.append({
"generation": gen,
"hash": arch_hash,
"fitness": best_fitness,
"config": best_candidate.__dict__,
})
logger.info(f"Gen {gen}: New best fitness={best_fitness:.4f}, hash={arch_hash}")
self.fitness_log.append(best_fitness)
return {
"best_fitness": best_fitness,
"best_config": best_config.__dict__ if best_config else None,
"generations_evolved": num_generations,
"architecture_history": self.architecture_history,
}
def self_reflect(self) -> Dict[str, Any]:
"""
Meta-cognitive reflection on system performance and architecture.
Inspired by GEA experience sharing and Yunjue Agent self-reflection.
"""
reflection = {
"generation": self.generation,
"total_architectures_tested": len(self.architecture_history),
"fitness_trend": self.fitness_log,
"memory_stats": {
"working_items": len(self.memory["working"].buffer),
"episodic_items": len(self.memory["temporal"].buffer),
},
"knowledge_stats": self.knowledge.stats(),
"agent_stats": self.agents.stats(),
"neuro_symbolic_balance": {
"symbolic_gate": torch.sigmoid(self.symbolic_gate).item(),
"neural_gate": torch.sigmoid(self.neural_gate).item(),
},
"recommendations": self._generate_recommendations(),
}
return reflection
def _generate_recommendations(self) -> List[str]:
"""Generate evolution directives based on performance analysis."""
recs = []
if len(self.fitness_log) > 5:
recent = self.fitness_log[-5:]
if max(recent) - min(recent) < 0.01:
recs.append("Fitness plateau detected. Increase mutation rate or population diversity.")
if recent[-1] < recent[0]:
recs.append("Performance declining. Consider rolling back to earlier architecture.")
sym_gate = torch.sigmoid(self.symbolic_gate).item()
if sym_gate < 0.3:
recs.append("Symbolic reasoning underutilized. Boost knowledge graph integration.")
elif sym_gate > 0.7:
recs.append("Symbolic dominance detected. Increase neural flexibility.")
return recs
def export_state(self) -> Dict[str, Any]:
"""Export full system state for checkpointing."""
return {
"config": self.config.__dict__,
"generation": self.generation,
"architecture_history": self.architecture_history,
"fitness_log": self.fitness_log,
"metadata": self.metadata,
"knowledge_state": self.knowledge.export(),
"memory_state": {
"working": self.memory["working"].export(),
"temporal": self.memory["temporal"].export(),
},
"model_state_dict": {k: v.cpu().tolist() for k, v in self.state_dict().items()},
}
@classmethod
def from_state(cls, state: Dict[str, Any]) -> "AetherCore":
"""Restore AETHER from checkpoint."""
config = AetherConfig(**state["config"])
core = cls(config=config, model_name=state["metadata"]["model_name"])
core.generation = state["generation"]
core.architecture_history = state["architecture_history"]
core.fitness_log = state["fitness_log"]
core.metadata = state["metadata"]
return core
|