File size: 8,701 Bytes
430bfd7 d4abd7f 430bfd7 51e44c2 430bfd7 d4abd7f 51e44c2 430bfd7 d4abd7f 1b634dd d4abd7f 430bfd7 d4abd7f 430bfd7 d4abd7f 430bfd7 51e44c2 430bfd7 a6e0d00 430bfd7 51e44c2 430bfd7 | 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 | """Three native-cell slice controls and cancellable, saved-frame playback."""
import asyncio
import json
import os
from pathlib import Path
from threading import Lock
from time import monotonic
import gradio as gr
import numpy as np
from matplotlib.backends.backend_agg import FigureCanvasAgg
from matplotlib.colors import LinearSegmentedColormap, SymLogNorm
from matplotlib.figure import Figure
try:
from .reader import PLANES, NativeDataset
except ImportError:
from reader import PLANES, NativeDataset
PLOT_LOCK = Lock()
CMAP = LinearSegmentedColormap.from_list(
"native-cyan", ["#06111d", "#0e415c", "#2199bd", "#a9e1df", "#fff3b0"]
)
CSS = """
.gradio-container {background:#06111d !important; color:#e9f1f5 !important; max-width:1400px !important}
.prose {color:#e9f1f5 !important} footer {display:none !important}
"""
def render_slice(dataset, frame, plane, coordinate, quantity, n):
vectors, _ = dataset.sample_plane(frame, plane, coordinate, quantity, n)
values = np.linalg.norm(vectors, axis=-1)
vmax = max(dataset.manifest["color_max"][quantity], 1e-15)
h, v, fixed = PLANES[plane]
with PLOT_LOCK:
figure = Figure(figsize=(4.4, 4.4), dpi=120, facecolor="#06111d")
canvas = FigureCanvasAgg(figure)
axes = figure.add_subplot(111, facecolor="#06111d")
picture = axes.imshow(
values.T,
origin="lower",
extent=(-1, 1, -1, 1),
cmap=CMAP,
norm=SymLogNorm(linthresh=vmax * 1e-3, vmin=0, vmax=vmax),
interpolation="nearest",
)
axes.set(
xlabel="xyz"[h],
ylabel="xyz"[v],
title=f"{plane[0]}–{plane[1]} · {'xyz'[fixed]} = {coordinate:.5f}",
)
axes.tick_params(colors="#9fb3c2")
for label in (axes.xaxis.label, axes.yaxis.label, axes.title):
label.set_color("#e9f1f5")
colorbar = figure.colorbar(picture, ax=axes, fraction=0.045, pad=0.035)
colorbar.ax.tick_params(colors="#9fb3c2")
figure.tight_layout()
canvas.draw()
return np.asarray(canvas.buffer_rgba())[..., :3].copy()
def render_state(dataset, frame, x, y, z, quantity, n):
if not np.isfinite(frame) or int(frame) != frame or int(n) != n:
raise ValueError("Choose a saved frame and supported display resolution")
frame, n = int(frame), int(n)
images = [
render_slice(dataset, frame, plane, coord, quantity, n)
for plane, coord in zip(("yz", "xz", "xy"), (x, y, z))
]
status = f"Frame {frame + 1}/{len(dataset.times)} · t = {dataset.times[frame]:.8f} · {quantity} · {n}² display samples per plane\n\nx = {x:.5f} · y = {y:.5f} · z = {z:.5f}\n\n{dataset.cache_status}"
return (*images, status)
def make_app(dataset):
with gr.Blocks(
title="Native Navier–Stokes explorer", delete_cache=(3600, 3600)
) as app:
gr.Markdown(
"## Explore the saved simulation\nMove the three fixed-coordinate planes, then play through saved time. Velocity and forcing are in model units."
)
if not dataset.manifest["complete_history"]:
gr.Markdown(
f"**Prototype subset:** {len(dataset.times)} of {dataset.manifest['source_frames']} saved states. This is not the complete animation."
)
with gr.Row():
quantity = gr.Radio(
["velocity", "force"], value="velocity", label="Quantity"
)
resolution = gr.Dropdown(
[128, 256, 512, 1024], value=256, label="Display samples per axis"
)
step = min(min(level["spacing"]) for level in dataset.levels)
with gr.Row():
x = gr.Slider(-1, 1, value=0, step=step / 2, label="x · YZ slice")
y = gr.Slider(-1, 1, value=0, step=step / 2, label="y · XZ slice")
z = gr.Slider(-1, 1, value=0, step=step / 2, label="z · XY slice")
with gr.Row():
images = [
gr.Image(
label=name,
interactive=False,
format="png",
type="numpy",
buttons=["download", "fullscreen"],
alt_text=f"Native {name} magnitude slice",
)
for name in ("YZ", "XZ", "XY")
]
time = gr.Slider(
0,
len(dataset.times) - 1,
value=0,
step=1,
label="t · saved frame",
info="Actual simulation time is shown below. Saved times are not uniformly spaced.",
)
with gr.Row():
previous = gr.Button("Previous frame")
play = gr.Button("Play", variant="primary")
pause = gr.Button("Pause")
following = gr.Button("Next frame")
status = gr.Markdown()
gr.Markdown(
"The viewer samples the finest native cell available at each location. Display resolution does not change solver resolution. No spatial or temporal interpolation, automatic peak tracking, or vector arrows. Playback targets five saved states per second and waits for data; it never skips a state.\n\nFinite manufactured-force experiment—not a singularity or convergence certificate."
)
inputs = [time, x, y, z, quantity, resolution]
outputs = [*images, status]
def render(*args):
return render_state(dataset, *args)
async def animate(frame, *args):
start = 0 if int(frame) == len(dataset.times) - 1 else int(frame)
for index in range(start, len(dataset.times)):
started = monotonic()
result = await asyncio.to_thread(render, index, *args)
yield index, *result
await asyncio.sleep(max(0, 0.2 - (monotonic() - started)))
playback = play.click(
animate,
inputs,
[time, *outputs],
api_name=False,
concurrency_id="native-playback",
concurrency_limit=4,
show_progress="hidden",
)
pause.click(fn=None, cancels=[playback], queue=False)
# One ordered event for all controls: a slower old request must not
# overwrite a newer quantity/coordinate choice. Playback uses a separate
# bounded group so one long animation cannot monopolize manual controls.
gr.on(
triggers=[
x.input,
y.input,
z.input,
time.input,
quantity.change,
resolution.change,
],
fn=render,
inputs=inputs,
outputs=outputs,
cancels=[playback],
trigger_mode="always_last",
concurrency_id="native-manual",
concurrency_limit=1,
show_progress="hidden",
)
for button, increment in ((previous, -1), (following, 1)):
def advance(frame, *args, amount=increment):
index = max(0, min(len(dataset.times) - 1, int(frame) + amount))
return index, *render(index, *args)
button.click(
advance,
inputs,
[time, *outputs],
cancels=[playback],
concurrency_id="native-manual",
concurrency_limit=1,
show_progress="hidden",
)
app.load(
render,
inputs,
outputs,
api_name="slice",
concurrency_id="native-manual",
concurrency_limit=1,
)
return app.queue(max_size=16, default_concurrency_limit=4)
if __name__ == "__main__":
config_path = Path(__file__).with_name("dataset.json")
config = json.loads(config_path.read_text()) if config_path.exists() else {}
dataset = NativeDataset(
folder=os.environ.get("NS_DATASET_LOCAL"),
repo_id=os.environ.get("NS_DATASET_REPO", config.get("repo_id")),
revision=os.environ.get("NS_DATASET_REVISION", config.get("revision")),
prefix=os.environ.get("NS_DATASET_PREFIX", config.get("prefix", "")),
)
if os.environ.get("SPACE_ID"):
dataset.warm_archive_cache()
make_app(dataset).launch(
server_name="0.0.0.0" if os.environ.get("SPACE_ID") else "127.0.0.1",
server_port=int(
os.environ.get("PORT", "7860" if os.environ.get("SPACE_ID") else "8877")
),
share=False,
max_threads=4,
theme=gr.themes.Base(primary_hue="cyan", neutral_hue="slate"),
css=CSS,
footer_links=[],
show_error=False,
)
|