Text Generation
PyTorch
English
French
hyperdimensional-computing
spiking-neural-networks
hdc
snn
lif
stdp
r-stdp
brain-inspired
cognitive-architecture
agentic
cpu-only
no-transformer
no-gpu
non-transformer
sparse-distributed-memory
kanerva
attractor-networks
global-workspace-theory
predictive-coding
neuromodulators
consciousness
kuramoto
vector-symbolic-architecture
vsa
one-shot-learning
instant-learning
pure-python
numpy
scipy
fastapi
web-dashboard
multi-modal
bpe
benchmark
beam-search
attention
reinforcement-learning
n-gram
kneser-ney
generative-ai
reasoning
creative-writing
research
prototype
File size: 14,105 Bytes
3636a69 | 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 | """
test_aether_v3.py — Test suite for AETHER v3 (multi-modal GPT killer).
Covers all 6 new v3 capabilities:
1. Semantic HD embeddings (pretrained.py) — categories, synonyms, antonyms
2. HD n-gram LM with backoff + temperature sampling + beam search
3. N-hop arbitrary depth with sub-goaling + path finding + reachability
4. Tool composer (hybrid HD + pattern matching)
5. HD RAG (web fetch + offline wiki fallback)
6. Multi-modal HD encoding (images + audio + cross-modal)
"""
from __future__ import annotations
import sys
import os
import time
import numpy as np
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from aether import (
AETHER,
SemanticKB,
HDLanguageModel,
ToolComposer, ComposedExecutor,
HDRAGStore, WebFetcher, get_global_rag, reset_global_rag,
ImageHDEncoder, AudioHDEncoder, CrossModalSpace,
integrate_into_agent,
)
def banner(title: str) -> None:
print()
print("=" * 76)
print(f" {title}")
print("=" * 76)
# --------------------------------------------------------------------------- #
# 1. Semantic embeddings
# --------------------------------------------------------------------------- #
def test_semantic_embeddings():
banner("1. Semantic HD Embeddings (categories + synonyms + antonyms)")
skb = SemanticKB(dim=4096)
skb.build()
print(f" Built semantic KB: {skb.stats()}")
print("\n Within-category similarity (HIGH):")
pairs = [("dog", "cat"), ("paris", "london"), ("red", "blue"), ("three", "seven")]
for w1, w2 in pairs:
print(f" sim({w1!r:10s}, {w2!r:10s}) = {skb.similarity(w1, w2):+.4f}")
print("\n Cross-category similarity (LOW):")
pairs = [("dog", "paris"), ("red", "seven"), ("bread", "happy")]
for w1, w2 in pairs:
print(f" sim({w1!r:10s}, {w2!r:10s}) = {skb.similarity(w1, w2):+.4f}")
print("\n Synonym similarity (~1.0):")
for w1, w2 in [("happy", "joyful"), ("big", "large"), ("smart", "intelligent")]:
print(f" sim({w1!r:12s}, {w2!r:12s}) = {skb.similarity(w1, w2):+.4f}")
print("\n Antonym similarity (LOW):")
for w1, w2 in [("hot", "cold"), ("big", "small"), ("day", "night")]:
print(f" sim({w1!r:12s}, {w2!r:12s}) = {skb.similarity(w1, w2):+.4f}")
print(f"\n Nearest neighbors of 'dog':")
for w, s in skb.nearest_neighbors("dog", top_k=3):
print(f" {w!r:15s} : {s:+.4f}")
# Test integration into agent
agent = AETHER()
before = len(agent.assoc.vocab)
added = integrate_into_agent(agent)
after = len(agent.assoc.vocab)
print(f"\n Integration into agent: {added} vectors added ({before} -> {after} vocab)")
print(" -> Semantic embeddings OK")
# --------------------------------------------------------------------------- #
# 2. HD n-gram LM
# --------------------------------------------------------------------------- #
def test_hd_lm():
banner("2. HD n-gram LM with backoff + sampling + beam search")
agent = AETHER()
lm = HDLanguageModel(agent.assoc, dim=agent.dim)
sentences = [
"the sun is a star",
"the moon is a satellite",
"water is a liquid",
"gold is a metal",
"python is a programming language",
"the cat is an animal",
"the dog is an animal",
"tokyo is the capital of japan",
"paris is the capital of france",
"ottawa is the capital of canada",
]
print(f" Training LM on {len(sentences)} sentences...")
for s in sentences:
lm.learn_sequence(s.lower().split())
print(f" LM stats: {lm.stats()}")
print("\n Prediction with backoff:")
contexts = [
["the", "sun"],
["the", "moon"],
["python", "is"],
["tokyo", "is", "the", "capital"],
]
for ctx in contexts:
pred = lm.predict_next(ctx)
if pred:
tok, sim, order = pred
print(f" ctx={ctx!r:45s} -> {tok!r:12s} (sim={sim:.3f}, order={order})")
print("\n Greedy generation (T=0):")
rng = np.random.default_rng(42)
for prompt in [["the", "sun"], ["python", "is"], ["paris", "is", "the"]]:
gen = lm.generate(prompt, max_tokens=8, temperature=0.0, top_k=3, rng=rng)
print(f" {' '.join(prompt):20s} ... -> {' '.join(gen)}")
print("\n Sampled generation (T=1.0, 3 samples):")
for prompt in [["the", "sun"]]:
for i in range(3):
rng_i = np.random.default_rng(i)
gen = lm.generate(prompt, max_tokens=8, temperature=1.0, top_k=5, rng=rng_i)
print(f" sample {i+1}: {' '.join(prompt)} ... -> {' '.join(gen)}")
print("\n Beam search (beam_width=3):")
for prompt in [["the", "sun"], ["python", "is"]]:
beams = lm.beam_search(prompt, max_tokens=6, beam_width=3)
print(f" prompt: {' '.join(prompt)}")
for seq, score in beams:
print(f" [{score:.3f}] {' '.join(seq)}")
print(" -> HD LM OK")
# --------------------------------------------------------------------------- #
# 3. N-hop with sub-goaling
# --------------------------------------------------------------------------- #
def test_n_hop_subgoaling():
banner("3. N-hop with sub-goaling + path finding + reachability")
agent = AETHER()
extra = [
"Montreal is located in Canada",
"Toronto is located in Canada",
"Canada is located in America",
"America is located in Earth",
"Lyon is located in France",
"France is located in Europe",
"Europe is located in Earth",
"Ottawa is the capital of Canada",
"Paris is the capital of France",
"Tokyo is the capital of Japan",
]
for f in extra:
agent.teach(f, silent=True)
print(f" KB: {len(agent.assoc.triples)} triples")
print("\n Direct lookup with explicit verification:")
proof = agent.inference.n_hop_with_subgoaling("Montreal", "located_in", max_depth=4)
print(f" {agent.inference.explain(proof)}")
print("\n Path finding (Montreal -> Earth):")
paths = agent.inference.find_paths("Montreal", "Earth", max_depth=5, max_paths=3)
for i, p in enumerate(paths):
print(f" Path {i+1}:")
for step in p.steps:
s, pr, o = step.conclusion
print(f" ({s}, {pr}, {o}) [conf={step.confidence:.3f}]")
print("\n Reachability from Montreal (3 hops):")
reachable = agent.inference.reachable("Montreal", max_hops=3)
for entity, (path, conf) in sorted(reachable.items(), key=lambda x: -x[1][1])[:6]:
print(f" {entity:15s} via {' -> '.join(path):40s} conf={conf:.3f}")
print("\n Multi-hop query (Montreal -> located_in -> capital_of):")
proof = agent.inference.multi_hop_query("Montreal", ["located_in", "capital_of"])
print(f" {agent.inference.explain(proof)}")
print(" -> N-hop sub-goaling OK")
# --------------------------------------------------------------------------- #
# 4. Tool composer
# --------------------------------------------------------------------------- #
def test_tool_composer():
banner("4. Tool Composer (hybrid HD + pattern matching)")
agent = AETHER()
composer = ToolComposer(agent.encoder, agent.tools, agent.inference)
executor = ComposedExecutor(agent)
for f in ["Tokyo is the capital of Japan", "Paris is the capital of France"]:
agent.teach(f, silent=True)
questions = [
"compare Paris and Tokyo",
"how many triples are in the KB",
"calculate 5 * 8 + 3",
"explain Python",
"what is the capital of France",
"where is Montreal located",
"list all facts",
"recall Paris",
"time",
"summarize 3",
]
print(f" Testing {len(questions)} questions:\n")
for q in questions:
nodes, rationale = composer.compose(q)
print(f" Q: {q}")
print(f" rationale: {rationale}")
for i, n in enumerate(nodes):
deps = f" depends_on={n.depends_on}" if n.depends_on else ""
print(f" node {i}: {n.tool_name}({n.args!r}){deps}")
final, _ = executor.execute(nodes)
print(f" => {str(final)[:80]}")
print()
print(" -> Tool composer OK")
# --------------------------------------------------------------------------- #
# 5. HD RAG
# --------------------------------------------------------------------------- #
def test_hd_rag():
banner("5. HD RAG (web fetch + offline wiki fallback)")
reset_global_rag()
agent = AETHER()
rag = get_global_rag(agent.encoder)
print(f" RAG store stats: {rag.stats()}")
print(f" Has internet: {WebFetcher(rag).has_internet()}")
print("\n Retrieval tests (should find the right topic):")
queries = ["tokyo", "paris", "python", "aether", "transformer",
"kanerva", "water", "sun", "japan", "france"]
correct = 0
for q in queries:
results = rag.retrieve(q, top_k=1)
if results:
doc, sim = results[0]
ok = q in doc.topic.lower() or doc.topic.lower() in q
if ok:
correct += 1
marker = "OK" if ok else "FAIL"
print(f" [{marker}] {q!r:15s} -> [{doc.topic}] (sim={sim:.3f})")
print(f"\n Retrieval accuracy: {correct}/{len(queries)}")
print("\n Tool-style fetch:")
from aether.tools import ToolContext
ctx = ToolContext(agent.encoder, agent.assoc)
from aether.web import tool_web_search, tool_rag_query, tool_rag_stats
print(f" web_search Tokyo: {tool_web_search('Tokyo', ctx)[:80]}...")
print(f" rag_query paris:")
for line in tool_rag_query('paris', ctx).split('\n'):
print(f" {line}")
print(f" {tool_rag_stats('', ctx)}")
print(" -> HD RAG OK")
# --------------------------------------------------------------------------- #
# 6. Multi-modal
# --------------------------------------------------------------------------- #
def test_multimodal():
banner("6. Multi-modal HD Encoding (images + audio + cross-modal)")
print(" === Image HD Encoder ===")
img_enc = ImageHDEncoder(dim=4096, grid_size=16, intensity_bins=8)
# Create test images
square = np.zeros((32, 32))
square[8:24, 8:24] = 1.0
circle = np.zeros((32, 32))
for i in range(32):
for j in range(32):
if (i - 16) ** 2 + (j - 16) ** 2 < 100:
circle[i, j] = 1.0
noise = np.random.rand(32, 32)
v_sq = img_enc.encode(square)
v_ci = img_enc.encode(circle)
v_no = img_enc.encode(noise)
v_sq2 = img_enc.encode(square.copy())
print(f" sim(square, square_copy) = {v_sq.similarity(v_sq2):.3f} (~1.0)")
print(f" sim(square, circle) = {v_sq.similarity(v_ci):.3f} (moderate)")
print(f" sim(square, noise) = {v_sq.similarity(v_no):.3f} (~0)")
print("\n === Audio HD Encoder ===")
aud_enc = AudioHDEncoder(dim=4096, n_frames=16, n_bands=16, energy_bins=8)
v_440 = aud_enc.encode_sine(440)
v_440b = aud_enc.encode_sine(440)
v_880 = aud_enc.encode_sine(880)
v_220 = aud_enc.encode_sine(220)
v_chord = aud_enc.encode_chord([440, 554, 659])
print(f" sim(440, 440_copy) = {v_440.similarity(v_440b):.3f} (~1.0)")
print(f" sim(440, 880) = {v_440.similarity(v_880):.3f} (moderate)")
print(f" sim(440, 220) = {v_440.similarity(v_220):.3f} (moderate)")
print(f" sim(440, chord) = {v_440.similarity(v_chord):.3f} (lower)")
print("\n === Cross-modal Space ===")
cms = CrossModalSpace(dim=4096)
cms.add_image_pattern([
" ",
" ########### ",
" ############### ",
" ################# ",
" ################### ",
" ################### ",
" ################### ",
" ################# ",
" ############### ",
" ########### ",
], label="circle_pattern")
cms.add_image_pattern([
" ",
" ################ ",
" ################ ",
" ################ ",
" ",
], label="rectangle_pattern")
cms.add_sine(440, label="A4_note")
cms.add_sine(880, label="A5_note")
print(f" CrossModal stats: {cms.stats()}")
results = cms.find_similar_image_pattern([
" ",
" ########### ",
" ############### ",
" ################# ",
" ################### ",
" ################### ",
" ################### ",
" ################# ",
" ############### ",
" ########### ",
], top_k=3)
print(f"\n Find similar to circle_pattern:")
for mod, label, sim in results:
print(f" [{mod}] {label!r:25s} sim={sim:.3f}")
results = cms.find_similar_sine(440, top_k=3)
print(f"\n Find similar to A4 (440 Hz):")
for mod, label, sim in results:
print(f" [{mod}] {label!r:25s} sim={sim:.3f}")
print(" -> Multi-modal OK")
# --------------------------------------------------------------------------- #
# main
# --------------------------------------------------------------------------- #
def main():
tests = [
test_semantic_embeddings,
test_hd_lm,
test_n_hop_subgoaling,
test_tool_composer,
test_hd_rag,
test_multimodal,
]
for t in tests:
try:
t()
except Exception as e:
import traceback
print(f"\n [FAILED: {t.__name__}: {e}]")
traceback.print_exc()
print("\n" + "=" * 76)
print(" All v3 tests complete.")
print("=" * 76)
if __name__ == "__main__":
main()
|