Upload folder using huggingface_hub

.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+*.pyc

README.md

@@ -5,6 +5,7 @@ colorFrom: purple
 colorTo: green
 sdk: gradio
 sdk_version: 4.44.0
+python_version: "3.12"
 app_file: app.py
 pinned: true
 license: apache-2.0

app.py

@@ -74,6 +74,14 @@ _CSS = """<style>
 .dcx .caret{display:inline-block;width:9px;height:17px;border-radius:2px;background:#7ad1ff;margin-left:2px;vertical-align:text-bottom;animation:dcxb .8s steps(1) infinite}
 @keyframes dcxb{50%{opacity:0}}
 .dcx .sub{color:#889;font-size:12px;line-height:1.5;margin-top:8px}
+.dcx .stats{display:flex;gap:10px;flex-wrap:wrap;margin:10px 0}
+.dcx .stat{flex:1;min-width:130px;text-align:center;background:#14141c;border:1px solid #2a2a35;border-radius:12px;padding:13px 8px}
+.dcx .stat .v{font-size:28px;font-weight:800;line-height:1}
+.dcx .stat .l{font-size:11px;color:#99a;margin-top:6px}
+.dcx table{border-collapse:collapse;width:100%;margin:8px 0;font-size:13.5px}
+.dcx th,.dcx td{border:1px solid #2a2a35;padding:7px 10px;text-align:left}
+.dcx th{background:#14141c;color:#bcd}
+.dcx td.n{text-align:right;font-variant-numeric:tabular-nums}
 </style>"""
 
 
@@ -168,6 +176,28 @@ def route_run(seq, n_bases, do_gen):
         yield _wrap(head + f"<div class='sub'>{FOOTER}</div>")
 
 
+STATS_HTML = _wrap(
+    "<div class='h'>📊 DaisyChain vs Carbon-500M — the fair baseline</div>"
+    "<div class='stats'>"
+    "<div class='stat'><div class='v' style='color:#37b24d'>94.8%</div>"
+    "<div class='l'>routing accuracy<br>(held-out)</div></div>"
+    "<div class='stat'><div class='v' style='color:#7c5cff'>≈295M</div>"
+    "<div class='l'>total params<br>(4 × ~74M) &lt; Carbon-500M</div></div>"
+    "<div class='stat'><div class='v' style='color:#22b8cf'>~7×</div>"
+    "<div class='l'>cheaper per query<br>(one 74M specialist active)</div></div>"
+    "</div>"
+    "<table><tr><th>metric</th><th>DaisyChain</th><th>Carbon-500M</th></tr>"
+    "<tr><td>Likelihood — bits/base (↓ better)</td><td class='n'>1.86</td><td class='n'>1.75</td></tr>"
+    "<tr><td>Seq-recovery, eukaryote (↑ better)</td><td class='n'>31.8%</td><td class='n'>42.2%</td></tr>"
+    "<tr><td>Seq-recovery, bacteria (↑ better)</td><td class='n'>34.0%</td><td class='n'>49.5%</td></tr>"
+    "</table>"
+    "<div class='sub'>Four ~74M specialists (≈295M total, <b>under Carbon-500M</b>); only one runs per "
+    "query, so it's ~7× cheaper per token. Behind the 500M / 1T-token monolith but within striking "
+    "distance — the gap is concentrated in the structured domains (mRNA, bacteria) and keeps closing "
+    "with more per-domain training. Same protocols as Carbon's eval suite (sequence recovery; per-base "
+    "likelihood). Carbon-500M is the right yardstick for a sub-500M modular set, not the 3B flagship.</div>")
+
+
 HERO = """# 🌼 DaisyChain — a modular genomic mind
 **Four ~74M DNA/RNA specialists (≈295M total, under Carbon-500M)** — 🧬 Eukaryote, 🦠 Prokaryote,
 📜 mRNA, ✂️ mRNA-splice — each **distilled per-domain from Carbon-500M**. A learned router reads how
@@ -198,6 +228,7 @@ def build():
                             label="Example sequences (one per domain)")
             except Exception:
                 pass
+            gr.HTML(STATS_HTML)
     return demo