Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,481 +1,315 @@
|
|
1 |
-
# app.py β
|
2 |
-
|
3 |
-
from dataclasses import dataclass, asdict
|
4 |
-
from typing import List, Tuple, Dict, Any, Optional
|
5 |
-
from functools import lru_cache
|
6 |
-
|
7 |
-
import numpy as np
|
8 |
-
import plotly.graph_objs as go
|
9 |
-
import plotly.io as pio
|
10 |
-
import gradio as gr
|
11 |
-
import pandas as pd
|
12 |
-
|
13 |
-
import torch
|
14 |
-
import torch.nn as nn
|
15 |
-
import torch.optim as optim
|
16 |
-
|
17 |
-
from data_utils import load_piqa, load_hellaswag, hash_vectorize
|
18 |
|
19 |
# =========================
|
20 |
-
#
|
21 |
# =========================
|
22 |
-
|
23 |
-
:root {
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
"""
|
37 |
|
38 |
-
#
|
39 |
-
# GENOME
|
40 |
-
# =========================
|
41 |
-
@dataclass
|
42 |
-
class Genome:
|
43 |
-
d_model: int
|
44 |
-
n_layers: int
|
45 |
-
n_heads: int
|
46 |
-
ffn_mult: float
|
47 |
-
memory_tokens: int
|
48 |
-
dropout: float
|
49 |
-
species: int = 0
|
50 |
-
fitness: float = float("inf")
|
51 |
-
acc: Optional[float] = None
|
52 |
-
|
53 |
-
def vector(self) -> np.ndarray:
|
54 |
-
return np.array([
|
55 |
-
self.d_model / 1024.0,
|
56 |
-
self.n_layers / 24.0,
|
57 |
-
self.n_heads / 32.0,
|
58 |
-
self.ffn_mult / 8.0,
|
59 |
-
self.memory_tokens / 64.0,
|
60 |
-
self.dropout / 0.5
|
61 |
-
], dtype=np.float32)
|
62 |
-
|
63 |
-
def random_genome(rng: random.Random) -> Genome:
|
64 |
-
return Genome(
|
65 |
-
d_model=rng.choice([256, 384, 512, 640]),
|
66 |
-
n_layers=rng.choice([4, 6, 8, 10, 12]),
|
67 |
-
n_heads=rng.choice([4, 6, 8, 10, 12]),
|
68 |
-
ffn_mult=rng.choice([2.0, 3.0, 4.0, 6.0]),
|
69 |
-
memory_tokens=rng.choice([0, 4, 8, 16]),
|
70 |
-
dropout=rng.choice([0.0, 0.05, 0.1, 0.15]),
|
71 |
-
species=rng.randrange(5)
|
72 |
-
)
|
73 |
-
|
74 |
-
def mutate(g: Genome, rng: random.Random, rate: float) -> Genome:
|
75 |
-
g = Genome(**asdict(g))
|
76 |
-
if rng.random() < rate: g.d_model = rng.choice([256, 384, 512, 640])
|
77 |
-
if rng.random() < rate: g.n_layers = rng.choice([4, 6, 8, 10, 12])
|
78 |
-
if rng.random() < rate: g.n_heads = rng.choice([4, 6, 8, 10, 12])
|
79 |
-
if rng.random() < rate: g.ffn_mult = rng.choice([2.0, 3.0, 4.0, 6.0])
|
80 |
-
if rng.random() < rate: g.memory_tokens = rng.choice([0, 4, 8, 16])
|
81 |
-
if rng.random() < rate: g.dropout = rng.choice([0.0, 0.05, 0.1, 0.15])
|
82 |
-
if rng.random() < rate * 0.5: g.species = rng.randrange(5)
|
83 |
-
g.fitness = float("inf"); g.acc = None
|
84 |
-
return g
|
85 |
-
|
86 |
-
def crossover(a: Genome, b: Genome, rng: random.Random) -> Genome:
|
87 |
-
return Genome(
|
88 |
-
d_model = a.d_model if rng.random()<0.5 else b.d_model,
|
89 |
-
n_layers = a.n_layers if rng.random()<0.5 else b.n_layers,
|
90 |
-
n_heads = a.n_heads if rng.random()<0.5 else b.n_heads,
|
91 |
-
ffn_mult = a.ffn_mult if rng.random()<0.5 else b.ffn_mult,
|
92 |
-
memory_tokens = a.memory_tokens if rng.random()<0.5 else b.memory_tokens,
|
93 |
-
dropout = a.dropout if rng.random()<0.5 else b.dropout,
|
94 |
-
species = a.species if rng.random()<0.5 else b.species,
|
95 |
-
fitness = float("inf"), acc=None
|
96 |
-
)
|
97 |
-
|
98 |
-
# =========================
|
99 |
-
# PROXY FITNESS
|
100 |
-
# =========================
|
101 |
-
def rastrigin(x: np.ndarray) -> float:
|
102 |
-
A, n = 10.0, x.shape[0]
|
103 |
-
return A * n + np.sum(x**2 - A * np.cos(2 * math.pi * x))
|
104 |
-
|
105 |
-
class TinyMLP(nn.Module):
|
106 |
-
def __init__(self, in_dim: int, genome: Genome):
|
107 |
-
super().__init__()
|
108 |
-
h1 = max(64, int(0.25 * genome.d_model))
|
109 |
-
h2 = max(32, int(genome.ffn_mult * 32))
|
110 |
-
self.net = nn.Sequential(
|
111 |
-
nn.Linear(in_dim, h1), nn.ReLU(),
|
112 |
-
nn.Linear(h1, h2), nn.ReLU(),
|
113 |
-
nn.Linear(h2, 1)
|
114 |
-
)
|
115 |
-
def forward(self, x): return self.net(x).squeeze(-1)
|
116 |
-
|
117 |
-
from functools import lru_cache
|
118 |
-
@lru_cache(maxsize=4)
|
119 |
-
def _cached_dataset(name: str):
|
120 |
-
try:
|
121 |
-
if name.startswith("PIQA"): return load_piqa(subset=800, seed=42)
|
122 |
-
if name.startswith("HellaSwag"): return load_hellaswag(subset=800, seed=42)
|
123 |
-
except Exception:
|
124 |
-
return None
|
125 |
-
return None
|
126 |
-
|
127 |
-
def _train_eval_proxy(genome: Genome, dataset_name: str, explore: float, device: str="cpu"):
|
128 |
-
data = _cached_dataset(dataset_name)
|
129 |
-
if data is None:
|
130 |
-
v = genome.vector() * 2 - 1
|
131 |
-
base = rastrigin(v)
|
132 |
-
parsimony = 0.001 * (genome.d_model + 50*genome.n_layers + 20*genome.n_heads + 100*genome.memory_tokens)
|
133 |
-
noise = np.random.normal(scale=0.05 * max(0.0, min(1.0, explore)))
|
134 |
-
return float(base + parsimony + noise), None
|
135 |
-
|
136 |
-
Xtr_txt, ytr, Xva_txt, yva = data
|
137 |
-
nfeat = 4096
|
138 |
-
Xtr = hash_vectorize(Xtr_txt, n_features=nfeat, seed=1234)
|
139 |
-
Xva = hash_vectorize(Xva_txt, n_features=nfeat, seed=5678)
|
140 |
-
|
141 |
-
Xtr_t = torch.from_numpy(Xtr); ytr_t = torch.from_numpy(ytr.astype(np.float32))
|
142 |
-
Xva_t = torch.from_numpy(Xva); yva_t = torch.from_numpy(yva.astype(np.float32))
|
143 |
-
|
144 |
-
model = TinyMLP(nfeat, genome).to(device)
|
145 |
-
opt = optim.AdamW(model.parameters(), lr=2e-3)
|
146 |
-
lossf = nn.BCEWithLogitsLoss()
|
147 |
-
|
148 |
-
model.train(); steps, bs, N = 120, 256, Xtr_t.size(0)
|
149 |
-
for _ in range(steps):
|
150 |
-
idx = torch.randint(0, N, (bs,))
|
151 |
-
xb = Xtr_t[idx].to(device); yb = ytr_t[idx].to(device)
|
152 |
-
logits = model(xb); loss = lossf(logits, yb)
|
153 |
-
opt.zero_grad(); loss.backward()
|
154 |
-
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
|
155 |
-
opt.step()
|
156 |
-
|
157 |
-
model.eval()
|
158 |
-
with torch.no_grad():
|
159 |
-
logits = model(Xva_t.to(device))
|
160 |
-
probs = torch.sigmoid(logits).cpu().numpy()
|
161 |
-
|
162 |
-
if dataset_name.startswith("PIQA"):
|
163 |
-
probs = probs.reshape(-1,2); yva2 = yva.reshape(-1,2)
|
164 |
-
pred = (probs[:,0] > probs[:,1]).astype(np.int64)
|
165 |
-
truth = (yva2[:,0] == 1).astype(np.int64)
|
166 |
-
acc = float((pred == truth).mean())
|
167 |
-
else:
|
168 |
-
probs = probs.reshape(-1,4); yva2 = yva.reshape(-1,4)
|
169 |
-
pred = probs.argmax(axis=1); truth = yva2.argmax(axis=1)
|
170 |
-
acc = float((pred == truth).mean())
|
171 |
-
|
172 |
-
parsimony = 0.00000002 * (genome.d_model**2 * genome.n_layers) + 0.0001 * genome.memory_tokens
|
173 |
-
noise = np.random.normal(scale=0.01 * max(0.0, min(1.0, explore)))
|
174 |
-
fitness = (1.0 - acc) + parsimony + noise
|
175 |
-
return float(max(0.0, min(1.5, fitness))), float(acc)
|
176 |
-
|
177 |
-
def evaluate_genome(genome: Genome, dataset: str, explore: float):
|
178 |
-
if dataset == "Demo (Surrogate)":
|
179 |
-
v = genome.vector() * 2 - 1
|
180 |
-
base = rastrigin(v)
|
181 |
-
parsimony = 0.001 * (genome.d_model + 50*genome.n_layers + 20*genome.n_heads + 100*genome.memory_tokens)
|
182 |
-
noise = np.random.normal(scale=0.05 * max(0.0, min(1.0, explore)))
|
183 |
-
return float(base + parsimony + noise), None
|
184 |
-
if dataset.startswith("PIQA"): return _train_eval_proxy(genome, "PIQA", explore)
|
185 |
-
if dataset.startswith("HellaSwag"): return _train_eval_proxy(genome, "HellaSwag", explore)
|
186 |
-
v = genome.vector() * 2 - 1
|
187 |
-
return float(rastrigin(v)), None
|
188 |
|
189 |
# =========================
|
190 |
-
#
|
191 |
-
# =========================
|
192 |
-
BG = "#0F1A24"
|
193 |
-
DOT = "#93C5FD" # soft blue dot
|
194 |
-
SPHERE = "#cbd5e1" # subtle sphere tint
|
195 |
-
|
196 |
-
def sphere_project(points: np.ndarray) -> np.ndarray:
|
197 |
-
rng = np.random.RandomState(42)
|
198 |
-
W = rng.normal(size=(points.shape[1], 3)).astype(np.float32)
|
199 |
-
Y = points @ W
|
200 |
-
norms = np.linalg.norm(Y, axis=1, keepdims=True) + 1e-8
|
201 |
-
return (Y / norms) * 1.22
|
202 |
-
|
203 |
-
def make_idle_sphere() -> go.Figure:
|
204 |
-
# empty scatter, only sphere
|
205 |
-
u = np.linspace(0, 2*np.pi, 72)
|
206 |
-
v = np.linspace(0, np.pi, 36)
|
207 |
-
r = 1.22
|
208 |
-
xs = r*np.outer(np.cos(u), np.sin(v))
|
209 |
-
ys = r*np.outer(np.sin(u), np.sin(v))
|
210 |
-
zs = r*np.outer(np.ones_like(u), np.cos(v))
|
211 |
-
sphere = go.Surface(x=xs, y=ys, z=zs, opacity=0.06, showscale=False,
|
212 |
-
colorscale=[[0, SPHERE],[1, SPHERE]], hoverinfo="skip")
|
213 |
-
layout = go.Layout(
|
214 |
-
paper_bgcolor=BG, plot_bgcolor=BG,
|
215 |
-
title="Architecture Sphere (idle)", titlefont=dict(color="#E5E7EB"),
|
216 |
-
scene=dict(xaxis=dict(visible=False), yaxis=dict(visible=False), zaxis=dict(visible=False), bgcolor=BG),
|
217 |
-
margin=dict(l=0, r=0, t=36, b=0), showlegend=False, height=720,
|
218 |
-
font=dict(family="Inter, Arial, sans-serif", size=14, color="#E5E7EB")
|
219 |
-
)
|
220 |
-
return go.Figure(data=[sphere], layout=layout)
|
221 |
-
|
222 |
-
def make_sphere_figure(points3d: np.ndarray, genomes: List[Genome], gen_idx: int) -> go.Figure:
|
223 |
-
# single-color dots for a sober look
|
224 |
-
custom = np.array([[g.d_model, g.n_layers, g.n_heads, g.ffn_mult, g.memory_tokens, g.dropout,
|
225 |
-
g.species, g.fitness, (g.acc if g.acc is not None else -1.0)]
|
226 |
-
for g in genomes], dtype=np.float32)
|
227 |
-
scatter = go.Scatter3d(
|
228 |
-
x=points3d[:,0], y=points3d[:,1], z=points3d[:,2],
|
229 |
-
mode='markers',
|
230 |
-
marker=dict(size=7.2, color=DOT, opacity=0.92),
|
231 |
-
customdata=custom,
|
232 |
-
hovertemplate=(
|
233 |
-
"<b>Genome</b><br>"
|
234 |
-
"d_model=%{customdata[0]:.0f} Β· layers=%{customdata[1]:.0f} Β· heads=%{customdata[2]:.0f}<br>"
|
235 |
-
"ffn_mult=%{customdata[3]:.1f} Β· mem=%{customdata[4]:.0f} Β· drop=%{customdata[5]:.2f}<br>"
|
236 |
-
"species=%{customdata[6]:.0f}<br>"
|
237 |
-
"fitness=%{customdata[7]:.4f}<br>"
|
238 |
-
"accuracy=%{customdata[8]:.3f}<extra></extra>"
|
239 |
-
)
|
240 |
-
)
|
241 |
-
idle = make_idle_sphere()
|
242 |
-
layout = idle.layout.update(title=f"Evo Architecture Sphere β Gen {gen_idx}")
|
243 |
-
fig = go.Figure(data=idle.data + (scatter,), layout=layout)
|
244 |
-
return fig
|
245 |
-
|
246 |
-
def make_history_figure(history: List[Tuple[int,float,float]], metric: str) -> go.Figure:
|
247 |
-
xs = [h[0] for h in history]
|
248 |
-
if metric == "Accuracy":
|
249 |
-
ys = [h[2] if (h[2] == h[2]) else None for h in history]
|
250 |
-
title, ylab = "Best Accuracy per Generation", "Accuracy"
|
251 |
-
else:
|
252 |
-
ys = [h[1] for h in history]
|
253 |
-
title, ylab = "Best Fitness per Generation", "Fitness (β better)"
|
254 |
-
fig = go.Figure(data=[go.Scatter(x=xs, y=ys, mode="lines+markers", line=dict(width=2), marker=dict(color=DOT))])
|
255 |
-
fig.update_layout(
|
256 |
-
paper_bgcolor=BG, plot_bgcolor=BG, font=dict(color="#E5E7EB"),
|
257 |
-
title=title, xaxis_title="Generation", yaxis_title=ylab,
|
258 |
-
margin=dict(l=30, r=10, t=36, b=30), height=340
|
259 |
-
)
|
260 |
-
fig.update_xaxes(gridcolor="#1f2b36"); fig.update_yaxes(gridcolor="#1f2b36")
|
261 |
-
return fig
|
262 |
-
|
263 |
-
def fig_to_html(fig: go.Figure) -> str:
|
264 |
-
return pio.to_html(fig, include_plotlyjs=True, full_html=False, config=dict(displaylogo=False))
|
265 |
-
|
266 |
-
def approx_params(g: Genome) -> int:
|
267 |
-
per_layer = (4.0 + 2.0 * float(g.ffn_mult)) * (g.d_model ** 2)
|
268 |
-
total = per_layer * g.n_layers + 1000 * g.memory_tokens
|
269 |
-
return int(total)
|
270 |
-
|
271 |
-
# =========================
|
272 |
-
# RUNNER
|
273 |
-
# =========================
|
274 |
-
class EvoRunner:
|
275 |
-
def __init__(self):
|
276 |
-
self.lock = threading.Lock()
|
277 |
-
self.running = False
|
278 |
-
self.stop_flag = False
|
279 |
-
self.state: Dict[str, Any] = {}
|
280 |
-
# seed the idle sphere immediately
|
281 |
-
idle = fig_to_html(make_idle_sphere())
|
282 |
-
self.state = {"sphere_html": idle, "history_html": fig_to_html(make_history_figure([], "Accuracy")),
|
283 |
-
"top": [], "best": {}, "gen": 0, "dataset": "Demo (Surrogate)", "metric": "Accuracy"}
|
284 |
-
|
285 |
-
def run(self, dataset, pop_size, generations, mutation_rate, explore, exploit, seed, pace_ms, metric_choice):
|
286 |
-
rng = random.Random(int(seed))
|
287 |
-
self.stop_flag = False
|
288 |
-
self.running = True
|
289 |
-
|
290 |
-
pop: List[Genome] = [random_genome(rng) for _ in range(pop_size)]
|
291 |
-
for g in pop:
|
292 |
-
fit, acc = evaluate_genome(g, dataset, explore)
|
293 |
-
g.fitness, g.acc = fit, acc
|
294 |
-
|
295 |
-
history: List[Tuple[int,float,float]] = []
|
296 |
-
|
297 |
-
for gen in range(1, generations+1):
|
298 |
-
if self.stop_flag: break
|
299 |
-
|
300 |
-
k = max(2, int(2 + exploit * 5))
|
301 |
-
parents = [min(rng.sample(pop, k=k), key=lambda x: x.fitness) for _ in range(pop_size)]
|
302 |
-
|
303 |
-
children = []
|
304 |
-
for i in range(0, pop_size, 2):
|
305 |
-
a = parents[i]; b = parents[(i+1) % pop_size]
|
306 |
-
child1 = mutate(crossover(a,b,rng), rng, mutation_rate)
|
307 |
-
child2 = mutate(crossover(b,a,rng), rng, mutation_rate)
|
308 |
-
children.extend([child1, child2])
|
309 |
-
children = children[:pop_size]
|
310 |
-
|
311 |
-
for c in children:
|
312 |
-
fit, acc = evaluate_genome(c, dataset, explore)
|
313 |
-
c.fitness, c.acc = fit, acc
|
314 |
-
|
315 |
-
elite_n = max(1, pop_size // 10)
|
316 |
-
elites = sorted(pop, key=lambda x: x.fitness)[:elite_n]
|
317 |
-
pop = sorted(children, key=lambda x: x.fitness)
|
318 |
-
pop[-elite_n:] = elites
|
319 |
-
|
320 |
-
best = min(pop, key=lambda x: x.fitness)
|
321 |
-
history.append((gen, best.fitness, (best.acc if best.acc is not None else float("nan"))))
|
322 |
-
|
323 |
-
P = np.stack([g.vector() for g in pop], axis=0)
|
324 |
-
P3 = sphere_project(P)
|
325 |
-
sphere_fig = make_sphere_figure(P3, pop, gen)
|
326 |
-
hist_fig = make_history_figure(history, metric_choice)
|
327 |
-
|
328 |
-
top = sorted(pop, key=lambda x: x.fitness)[: min(12, len(pop))]
|
329 |
-
top_table = [{
|
330 |
-
"gen": gen, "fitness": round(t.fitness, 4),
|
331 |
-
"accuracy": (None if t.acc is None else round(float(t.acc), 4)),
|
332 |
-
"d_model": t.d_model, "layers": t.n_layers, "heads": t.n_heads,
|
333 |
-
"ffn_mult": t.ffn_mult, "mem": t.memory_tokens, "dropout": t.dropout,
|
334 |
-
"params_approx": approx_params(t)
|
335 |
-
} for t in top]
|
336 |
-
best_card = top_table[0] if top_table else {}
|
337 |
-
|
338 |
-
with self.lock:
|
339 |
-
self.state = {
|
340 |
-
"sphere_html": fig_to_html(sphere_fig),
|
341 |
-
"history_html": fig_to_html(hist_fig),
|
342 |
-
"top": top_table,
|
343 |
-
"best": best_card,
|
344 |
-
"gen": gen,
|
345 |
-
"dataset": dataset,
|
346 |
-
"metric": metric_choice
|
347 |
-
}
|
348 |
-
|
349 |
-
time.sleep(max(0.0, pace_ms/1000.0))
|
350 |
-
self.running = False
|
351 |
-
|
352 |
-
def start(self, *args, **kwargs):
|
353 |
-
if self.running: return
|
354 |
-
t = threading.Thread(target=self.run, args=args, kwargs=kwargs, daemon=True)
|
355 |
-
t.start()
|
356 |
-
|
357 |
-
def stop(self): self.stop_flag = True
|
358 |
-
|
359 |
-
def clear(self):
|
360 |
-
# stop and reset to idle sphere
|
361 |
-
self.stop_flag = True
|
362 |
-
idle = fig_to_html(make_idle_sphere())
|
363 |
-
with self.lock:
|
364 |
-
self.running = False
|
365 |
-
self.state = {"sphere_html": idle, "history_html": fig_to_html(make_history_figure([], "Accuracy")),
|
366 |
-
"top": [], "best": {}, "gen": 0, "dataset": "Demo (Surrogate)", "metric": "Accuracy"}
|
367 |
-
|
368 |
-
runner = EvoRunner()
|
369 |
-
|
370 |
# =========================
|
371 |
-
|
372 |
-
#
|
373 |
-
def start_evo(dataset, pop, gens, mut, explore, exploit, seed, pace_ms, metric_choice):
|
374 |
-
runner.start(dataset, int(pop), int(gens), float(mut), float(explore), float(exploit), int(seed), int(pace_ms), metric_choice)
|
375 |
-
return (gr.update(interactive=False), gr.update(interactive=True), gr.update(interactive=False))
|
376 |
-
|
377 |
-
def stop_evo():
|
378 |
-
runner.stop()
|
379 |
-
return (gr.update(interactive=True), gr.update(interactive=False), gr.update(interactive=True))
|
380 |
-
|
381 |
-
def clear_evo():
|
382 |
-
runner.clear()
|
383 |
-
# return updated visuals + reset buttons
|
384 |
-
sphere_html, history_html, stats_md, df = poll_state()
|
385 |
-
return sphere_html, history_html, stats_md, df, gr.update(interactive=True), gr.update(interactive=False), gr.update(interactive=True)
|
386 |
-
|
387 |
-
def poll_state():
|
388 |
-
with runner.lock:
|
389 |
-
s = runner.state.copy()
|
390 |
-
sphere_html = s.get("sphere_html", fig_to_html(make_idle_sphere()))
|
391 |
-
history_html = s.get("history_html", fig_to_html(make_history_figure([], "Accuracy")))
|
392 |
-
best = s.get("best", {})
|
393 |
-
gen = s.get("gen", 0)
|
394 |
-
dataset = s.get("dataset", "Demo (Surrogate)")
|
395 |
-
top = s.get("top", [])
|
396 |
-
if best:
|
397 |
-
acc_txt = "β" if best.get("accuracy") is None else f"{best.get('accuracy'):.3f}"
|
398 |
-
stats_md = (
|
399 |
-
f"**Dataset:** {dataset} \n"
|
400 |
-
f"**Generation:** {gen} \n"
|
401 |
-
f"**Best fitness:** {best.get('fitness','β')} \n"
|
402 |
-
f"**Best accuracy:** {acc_txt} \n"
|
403 |
-
f"**Config:** d_model={best.get('d_model')} Β· layers={best.get('layers')} Β· "
|
404 |
-
f"heads={best.get('heads')} Β· ffn_mult={best.get('ffn_mult')} Β· mem={best.get('mem')} Β· "
|
405 |
-
f"dropout={best.get('dropout')} \n"
|
406 |
-
f"**~Params (rough):** {best.get('params_approx'):,}"
|
407 |
-
)
|
408 |
-
else:
|
409 |
-
stats_md = "Ready. Press **Start** to evolve, or **Clear** anytime."
|
410 |
-
df = pd.DataFrame(top)
|
411 |
-
return sphere_html, history_html, stats_md, df
|
412 |
-
|
413 |
-
def export_snapshot():
|
414 |
-
from json import dumps
|
415 |
-
with runner.lock:
|
416 |
-
payload = dumps(runner.state, default=lambda o: o, indent=2)
|
417 |
-
path = "evo_snapshot.json"
|
418 |
-
with open(path, "w", encoding="utf-8") as f:
|
419 |
-
f.write(payload)
|
420 |
-
return path
|
421 |
-
|
422 |
-
# =========================
|
423 |
-
# BUILD UI
|
424 |
-
# =========================
|
425 |
-
with gr.Blocks(css=CUSTOM_CSS) as demo:
|
426 |
with gr.Column(elem_id="header"):
|
427 |
-
gr.Markdown("
|
428 |
-
|
|
|
|
|
429 |
with gr.Row():
|
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 |
with gr.Group(elem_classes=["panel"]):
|
457 |
-
|
458 |
-
|
459 |
-
|
|
|
|
|
|
|
460 |
with gr.Column(scale=2):
|
461 |
-
|
|
|
|
|
|
|
|
|
|
|
462 |
sphere_html = gr.HTML()
|
463 |
-
|
|
|
|
|
|
|
|
|
464 |
hist_html = gr.HTML()
|
|
|
|
|
465 |
with gr.Group(elem_classes=["panel"]):
|
466 |
-
|
467 |
-
|
468 |
-
|
469 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
470 |
stop.click(stop_evo, [], [start, stop, clear])
|
471 |
-
clear.click(
|
|
|
|
|
|
|
|
|
472 |
export_btn.click(export_snapshot, [], [export_file])
|
473 |
-
|
474 |
-
#
|
475 |
-
demo.load(poll_state, None, [sphere_html, hist_html, stats_md, top_df])
|
476 |
-
gr.Timer(0.7).tick(poll_state, None, [sphere_html, hist_html, stats_md, top_df])
|
477 |
|
478 |
if __name__ == "__main__":
|
479 |
-
demo.launch()
|
480 |
-
|
481 |
-
##
|
|
|
1 |
+
# app.py β Enhanced UI with better layout, visual hierarchy, and UX
|
2 |
+
# ... [All your imports and backend code remain the same] ...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
# =========================
|
5 |
+
# ENHANCED CSS
|
6 |
# =========================
|
7 |
+
ENHANCED_CSS = """
|
8 |
+
:root {
|
9 |
+
--radius: 14px;
|
10 |
+
--fg: #E5E7EB;
|
11 |
+
--muted: #94A3B8;
|
12 |
+
--line: #1f2b36;
|
13 |
+
--bg: #0F1A24;
|
14 |
+
--panel-bg: #0c161f;
|
15 |
+
--accent: #3B82F6;
|
16 |
+
--accent-hover: #2563EB;
|
17 |
+
--danger: #EF4444;
|
18 |
+
}
|
19 |
+
|
20 |
+
.gradio-container {
|
21 |
+
max-width: 1400px !important;
|
22 |
+
background: var(--bg);
|
23 |
+
padding: 16px !important;
|
24 |
+
}
|
25 |
+
|
26 |
+
#header {
|
27 |
+
padding: 16px 0;
|
28 |
+
margin-bottom: 16px;
|
29 |
+
border-bottom: 1px solid var(--line);
|
30 |
+
}
|
31 |
+
|
32 |
+
h1, h2, h3, .gr-markdown {
|
33 |
+
color: var(--fg);
|
34 |
+
}
|
35 |
+
|
36 |
+
.gr-button {
|
37 |
+
border-radius: 8px;
|
38 |
+
padding: 8px 16px;
|
39 |
+
transition: all 0.2s ease;
|
40 |
+
font-weight: 500 !important;
|
41 |
+
}
|
42 |
+
|
43 |
+
.btn-primary {
|
44 |
+
background: var(--accent) !important;
|
45 |
+
border: 1px solid var(--accent) !important;
|
46 |
+
}
|
47 |
+
|
48 |
+
.btn-primary:hover {
|
49 |
+
background: var(--accent-hover) !important;
|
50 |
+
}
|
51 |
+
|
52 |
+
.btn-secondary {
|
53 |
+
background: transparent !important;
|
54 |
+
border: 1px solid var(--line) !important;
|
55 |
+
}
|
56 |
+
|
57 |
+
.btn-danger {
|
58 |
+
background: var(--danger) !important;
|
59 |
+
border: 1px solid var(--danger) !important;
|
60 |
+
}
|
61 |
+
|
62 |
+
.control-group {
|
63 |
+
border: 1px solid var(--line);
|
64 |
+
border-radius: var(--radius);
|
65 |
+
background: var(--panel-bg);
|
66 |
+
padding: 20px;
|
67 |
+
margin-bottom: 20px;
|
68 |
+
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.05);
|
69 |
+
}
|
70 |
+
|
71 |
+
.panel {
|
72 |
+
border: 1px solid var(--line);
|
73 |
+
border-radius: var(--radius);
|
74 |
+
background: var(--panel-bg);
|
75 |
+
padding: 20px;
|
76 |
+
margin-bottom: 20px;
|
77 |
+
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.05);
|
78 |
+
}
|
79 |
+
|
80 |
+
.stats-panel {
|
81 |
+
background: linear-gradient(145deg, #0a121b, #0c161f);
|
82 |
+
border-left: 3px solid var(--accent);
|
83 |
+
}
|
84 |
+
|
85 |
+
#stats {
|
86 |
+
color: var(--fg);
|
87 |
+
line-height: 1.6;
|
88 |
+
}
|
89 |
+
|
90 |
+
#stats strong {
|
91 |
+
font-weight: 500;
|
92 |
+
color: var(--accent);
|
93 |
+
}
|
94 |
+
|
95 |
+
.param-slider {
|
96 |
+
margin-bottom: 12px;
|
97 |
+
}
|
98 |
+
|
99 |
+
.visualization-container {
|
100 |
+
display: flex;
|
101 |
+
flex-direction: column;
|
102 |
+
gap: 20px;
|
103 |
+
height: 100%;
|
104 |
+
}
|
105 |
+
|
106 |
+
.viz-panel {
|
107 |
+
flex: 1;
|
108 |
+
min-height: 300px;
|
109 |
+
}
|
110 |
+
|
111 |
+
.viz-header {
|
112 |
+
display: flex;
|
113 |
+
justify-content: space-between;
|
114 |
+
align-items: center;
|
115 |
+
margin-bottom: 12px;
|
116 |
+
padding-bottom: 8px;
|
117 |
+
border-bottom: 1px solid var(--line);
|
118 |
+
}
|
119 |
+
|
120 |
+
.viz-title {
|
121 |
+
font-size: 1.1rem;
|
122 |
+
font-weight: 500;
|
123 |
+
color: var(--accent);
|
124 |
+
}
|
125 |
+
|
126 |
+
.gen-counter {
|
127 |
+
font-size: 0.9rem;
|
128 |
+
background: rgba(59, 130, 246, 0.15);
|
129 |
+
padding: 4px 10px;
|
130 |
+
border-radius: 12px;
|
131 |
+
}
|
132 |
+
|
133 |
+
.slider-info {
|
134 |
+
display: flex;
|
135 |
+
justify-content: space-between;
|
136 |
+
font-size: 0.85rem;
|
137 |
+
color: var(--muted);
|
138 |
+
margin-top: 4px;
|
139 |
+
}
|
140 |
+
|
141 |
+
.controls-grid {
|
142 |
+
display: grid;
|
143 |
+
grid-template-columns: 1fr 1fr;
|
144 |
+
gap: 16px;
|
145 |
+
}
|
146 |
+
|
147 |
+
@media (max-width: 1200px) {
|
148 |
+
.controls-grid {
|
149 |
+
grid-template-columns: 1fr;
|
150 |
+
}
|
151 |
+
}
|
152 |
+
|
153 |
+
.data-table {
|
154 |
+
max-height: 400px;
|
155 |
+
overflow-y: auto;
|
156 |
+
}
|
157 |
+
|
158 |
+
.data-table table {
|
159 |
+
width: 100%;
|
160 |
+
border-collapse: collapse;
|
161 |
+
}
|
162 |
+
|
163 |
+
.data-table th {
|
164 |
+
background: rgba(15, 26, 36, 0.8);
|
165 |
+
position: sticky;
|
166 |
+
top: 0;
|
167 |
+
text-align: left;
|
168 |
+
padding: 10px 12px;
|
169 |
+
font-weight: 500;
|
170 |
+
color: var(--accent);
|
171 |
+
border-bottom: 1px solid var(--line);
|
172 |
+
}
|
173 |
+
|
174 |
+
.data-table td {
|
175 |
+
padding: 8px 12px;
|
176 |
+
border-bottom: 1px solid rgba(31, 43, 54, 0.5);
|
177 |
+
}
|
178 |
+
|
179 |
+
.data-table tr:hover {
|
180 |
+
background: rgba(31, 43, 54, 0.3);
|
181 |
+
}
|
182 |
+
|
183 |
+
.action-buttons {
|
184 |
+
display: flex;
|
185 |
+
gap: 12px;
|
186 |
+
margin-top: 20px;
|
187 |
+
}
|
188 |
+
|
189 |
+
.footer {
|
190 |
+
margin-top: 20px;
|
191 |
+
padding-top: 20px;
|
192 |
+
border-top: 1px solid var(--line);
|
193 |
+
font-size: 0.85rem;
|
194 |
+
color: var(--muted);
|
195 |
+
text-align: center;
|
196 |
+
}
|
197 |
"""
|
198 |
|
199 |
+
# ... [All your backend code remains the same] ...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
200 |
|
201 |
# =========================
|
202 |
+
# BUILD ENHANCED UI
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
203 |
# =========================
|
204 |
+
with gr.Blocks(css=ENHANCED_CSS, theme=gr.themes.Default()) as demo:
|
205 |
+
# Header
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
206 |
with gr.Column(elem_id="header"):
|
207 |
+
gr.Markdown("## 𧬠Neuroevolution Playground", elem_classes=["header-title"])
|
208 |
+
gr.Markdown("Evolve neural architectures using genetic algorithms",
|
209 |
+
elem_classes=["header-subtitle"])
|
210 |
+
|
211 |
with gr.Row():
|
212 |
+
# Left Panel - Controls
|
213 |
+
with gr.Column(scale=1):
|
214 |
+
# Parameters Group
|
215 |
+
with gr.Group(elem_classes=["control-group"]):
|
216 |
+
gr.Markdown("### π Evolution Parameters")
|
217 |
+
|
218 |
+
with gr.Column():
|
219 |
+
dataset = gr.Dropdown(
|
220 |
+
label="Evaluation Dataset",
|
221 |
+
choices=["Demo (Surrogate)", "PIQA (Phase 2)", "HellaSwag (Phase 2)"],
|
222 |
+
value="Demo (Surrogate)",
|
223 |
+
info="Dataset used for fitness evaluation"
|
224 |
+
)
|
225 |
+
|
226 |
+
with gr.Row():
|
227 |
+
with gr.Column():
|
228 |
+
pop = gr.Slider(8, 80, value=24, step=2, label="Population Size",
|
229 |
+
elem_classes=["param-slider"])
|
230 |
+
gens = gr.Slider(5, 200, value=60, step=1, label="Max Generations",
|
231 |
+
elem_classes=["param-slider"])
|
232 |
+
mut = gr.Slider(0.05, 0.9, value=0.25, step=0.01, label="Mutation Rate",
|
233 |
+
elem_classes=["param-slider"])
|
234 |
+
with gr.Column():
|
235 |
+
explore = gr.Slider(0.0, 1.0, value=0.35, step=0.05, label="Exploration",
|
236 |
+
elem_classes=["param-slider"])
|
237 |
+
exploit = gr.Slider(0.0, 1.0, value=0.65, step=0.05, label="Exploitation",
|
238 |
+
elem_classes=["param-slider"])
|
239 |
+
seed = gr.Number(value=42, label="Random Seed", precision=0)
|
240 |
+
|
241 |
+
pace = gr.Slider(0, 1000, value=120, step=10, label="Simulation Speed (ms)",
|
242 |
+
elem_classes=["param-slider"])
|
243 |
+
metric_choice = gr.Radio(choices=["Accuracy", "Fitness"], value="Accuracy",
|
244 |
+
label="History Metric Display")
|
245 |
+
|
246 |
+
# Status Panel
|
247 |
+
with gr.Group(elem_classes=["panel", "stats-panel"]):
|
248 |
+
gr.Markdown("### π Current Status")
|
249 |
+
stats_md = gr.Markdown("Ready. Press **Start** to begin evolution.", elem_id="stats")
|
250 |
+
|
251 |
+
# Action Buttons
|
252 |
+
with gr.Row(elem_classes=["action-buttons"]):
|
253 |
+
start = gr.Button("βΆ Start Evolution", variant="primary", elem_classes=["btn-primary"])
|
254 |
+
stop = gr.Button("βΉ Stop", variant="stop", elem_classes=["btn-danger"], interactive=False)
|
255 |
+
clear = gr.Button("β» Reset", elem_classes=["btn-secondary"])
|
256 |
+
|
257 |
+
# Export
|
258 |
with gr.Group(elem_classes=["panel"]):
|
259 |
+
gr.Markdown("### πΎ Export Results")
|
260 |
+
with gr.Row():
|
261 |
+
export_btn = gr.Button("Save Snapshot (JSON)")
|
262 |
+
export_file = gr.File(label="Download snapshot", visible=False)
|
263 |
+
|
264 |
+
# Right Panel - Visualizations
|
265 |
with gr.Column(scale=2):
|
266 |
+
# 3D Visualization
|
267 |
+
with gr.Group(elem_classes=["panel", "viz-panel"]):
|
268 |
+
with gr.Column(elem_classes=["viz-header"]):
|
269 |
+
with gr.Row():
|
270 |
+
gr.Markdown("### π Architecture Space", elem_classes=["viz-title"])
|
271 |
+
gen_counter = gr.Markdown("", elem_classes=["gen-counter"])
|
272 |
sphere_html = gr.HTML()
|
273 |
+
|
274 |
+
# History Visualization
|
275 |
+
with gr.Group(elem_classes=["panel", "viz-panel"]):
|
276 |
+
with gr.Column(elem_classes=["viz-header"]):
|
277 |
+
gr.Markdown("### π Performance History", elem_classes=["viz-title"])
|
278 |
hist_html = gr.HTML()
|
279 |
+
|
280 |
+
# Results Table
|
281 |
with gr.Group(elem_classes=["panel"]):
|
282 |
+
gr.Markdown("### π Top Genomes")
|
283 |
+
with gr.Column(elem_classes=["data-table"]):
|
284 |
+
top_df = gr.Dataframe(
|
285 |
+
label="",
|
286 |
+
headers=["Fitness", "Accuracy", "d_model", "Layers", "Heads", "FFN", "Mem", "Dropout", "Params"],
|
287 |
+
datatype=["number", "number", "number", "number", "number", "number", "number", "number", "number"],
|
288 |
+
wrap=True,
|
289 |
+
interactive=False
|
290 |
+
)
|
291 |
+
|
292 |
+
# Footer
|
293 |
+
with gr.Column(elem_classes=["footer"]):
|
294 |
+
gr.Markdown("Evotransformer Playground v1.0 β’ Using Plotly and Gradio")
|
295 |
+
|
296 |
+
# Wiring
|
297 |
+
start.click(
|
298 |
+
start_evo,
|
299 |
+
[dataset, pop, gens, mut, explore, exploit, seed, pace, metric_choice],
|
300 |
+
[start, stop, clear]
|
301 |
+
)
|
302 |
stop.click(stop_evo, [], [start, stop, clear])
|
303 |
+
clear.click(
|
304 |
+
clear_evo,
|
305 |
+
[],
|
306 |
+
[sphere_html, hist_html, stats_md, top_df, start, stop, clear]
|
307 |
+
)
|
308 |
export_btn.click(export_snapshot, [], [export_file])
|
309 |
+
|
310 |
+
# State polling
|
311 |
+
demo.load(poll_state, None, [sphere_html, hist_html, stats_md, top_df, gen_counter])
|
312 |
+
gr.Timer(0.7).tick(poll_state, None, [sphere_html, hist_html, stats_md, top_df, gen_counter])
|
313 |
|
314 |
if __name__ == "__main__":
|
315 |
+
demo.launch()
|
|
|
|