Spaces:
Running
Running
Download static/3d/surfaces/specdecode.js from SZLHOLDINGS/a11oy: direct link, hf CLI and curl.
- Browser
- Download file 19.9 kB
-
https://huggingface.co/spaces/SZLHOLDINGS/a11oy/resolve/e345f85fd97239d4814b1cacff13a2907f62f08a/static/3d/surfaces/specdecode.js
- Command line
-
hf download hf://spaces/SZLHOLDINGS/a11oy@e345f85fd97239d4814b1cacff13a2907f62f08a/static/3d/surfaces/specdecode.js
-
curl -L -o specdecode.js https://huggingface.co/spaces/SZLHOLDINGS/a11oy/resolve/e345f85fd97239d4814b1cacff13a2907f62f08a/static/3d/surfaces/specdecode.js
19.9 kB
| // SPDX-License-Identifier: Apache-2.0 | |
| // © 2026 Lutar, Stephen P. — SZL Holdings · ORCID 0009-0001-0110-4173 · Doctrine v11 | |
| // | |
| // surfaces/specdecode.js — SELF-SPECULATIVE MULTI-TOKEN DECODING organ for the | |
| // holographic frontier ring (Medusa/EAGLE-style draft-then-verify). Renders the | |
| // draft model proposing `draft_len` candidate tokens per step as a 3D token-tree | |
| // / pipeline, then colors each proposed token ACCEPTED (proof-teal) or REJECTED | |
| // (grey) per the live snapshot from /api/killinchu/v1/specdecode/simulate. A HUD | |
| // shows acceptance_rate + speedup_factor. Honesty label "MODELED" is read | |
| // VERBATIM from the JSON and displayed as-is; it is never upgraded. | |
| // | |
| // Surface export shape (mirrors testtime.js / interpretability.js exactly): | |
| // export default { id, title, endpoints, mount(ctx), unmount() } | |
| // ctx = { stage, container, live, label, THREE, szl3d } | |
| // | |
| // DATA SHOWN (all from live endpoint): | |
| // draft_len, trials, acceptance_rate, mean_accept_len, mean_tokens_per_step, | |
| // speedup_factor, per_step_accept_lengths[], lossless | |
| // | |
| // LEADERS ADOPTED & CITED (clean-room; NOT claimed as SZL's own): | |
| // Speculative Decoding (accept/reject arithmetic simulated here): | |
| // Leviathan, Kalman & Matias 2023, arXiv:2211.17192 | |
| // https://arxiv.org/abs/2211.17192 | |
| // Medusa (multiple decoding heads, tree verification — reference only): | |
| // Cai et al. 2024, github.com/FasterDecoding/Medusa | |
| // https://github.com/FasterDecoding/Medusa | |
| // EAGLE (feature-level extrapolation drafting — reference only): | |
| // Li et al. 2024, arXiv:2401.15077 | |
| // https://arxiv.org/abs/2401.15077 | |
| // | |
| // HONESTY LABELS: MODELED (deterministic re-implementation of the accept/reject | |
| // arithmetic; NOT Medusa/EAGLE; NEVER-CLAIMED-AS a production drafter). Read | |
| // verbatim from JSON; never upgraded here. | |
| // COLOURS: lattice-blue 0x5b8dee (draft tokens / pipeline), proof-teal 0x3af4c8 | |
| // (accepted tokens / HUD accent), greys (rejected tokens / degraded state). | |
| // Purple BANNED as UI/background. | |
| // 0 RUNTIME CDN. Vendored three.js r170 via page importmap. | |
| // DOCTRINE v11: degrades gracefully (grey) on 404/error; honesty label still shown. | |
| // Nothing here is in the locked-8. Λ stays Conjecture 1. Trust never 100%. | |
| const ID = "specdecode"; | |
| const TITLE = "Self-Speculative Decoding · Draft-then-Verify (live)"; | |
| // Endpoint is hosted on the dedicated killinchu Space (isolated compute), reached | |
| // cross-origin (killinchu returns access-control-allow-origin: https://a-11-oy.com). | |
| // This keeps the spec-decode organ's rebuilds/faults isolated from the flagship. | |
| const EP = "https://szlholdings-killinchu.hf.space/api/killinchu/v1/specdecode/simulate?seed=42&draft_len=6&trials=256"; | |
| // data-viz hues — purple BANNED | |
| const C_DRAFT = 0x5b8dee; // lattice-blue (draft-proposed token / pipeline spine) | |
| const C_ACCEPT = 0x3af4c8; // proof-teal (accepted token / HUD accent) | |
| const C_REJECT = 0x5a6570; // grey (rejected token) | |
| const C_DIM = 0x42505d; // grey (degraded / no-live-data) | |
| const C_GRID = 0x1b3a44; // floor / link colour | |
| // token-tree pipeline layout geometry | |
| const STEP_LEN = 2.1; // world-units between steps along X (pipeline axis) | |
| const TOKEN_GAP = 0.85; // world-units between tokens within a step (Y, draft depth) | |
| const MAX_STEPS = 10; // number of recent steps rendered along the pipeline | |
| const MAX_TOKENS = 12; // cap on draft_len rendered per step (perf) | |
| let _stage = null, _THREE = null, _ctx = null, _group = null, _overlay = null; | |
| let _frameReg = false, _polls = [], _el = {}, _badge = null; | |
| let _plain = false; | |
| // geometry handles | |
| let _floor = null; | |
| let _spine = null; // THREE.Line — pipeline spine (draft -> verify axis) | |
| let _tokenMesh = []; // Array<Array<THREE.Mesh>> — [step][tokenIdx] | |
| let _stepLinks = []; // Array<THREE.Line> — link from spine to each step's tokens | |
| let _marker = null; // THREE.Mesh — HUD "current step" pulsing marker | |
| // live state | |
| const S = { | |
| label: null, | |
| draftLen: null, // draft_len | |
| trials: null, // trials | |
| acceptRate: null, // acceptance_rate | |
| meanAccept: null, // mean_accept_len | |
| meanTokens: null, // mean_tokens_per_step | |
| speedup: null, // speedup_factor | |
| stepLengths: null, // per_step_accept_lengths[] | |
| lossless: null, // lossless (bool) | |
| state: "init", | |
| }; | |
| // ============================================================================= | |
| // mount(ctx) | |
| // ============================================================================= | |
| export function mount(ctx) { | |
| _ctx = ctx; _stage = ctx.stage; _THREE = ctx.THREE; | |
| _group = new _THREE.Group(); | |
| _stage.scene.add(_group); | |
| _stage.camera.position.set(4, 7, 18); | |
| try { if (_stage.controls && _stage.controls.target) { _stage.controls.target.set(6, 2, 0); _stage.controls.update(); } } catch (_) {} | |
| try { _stage.setBloom(true); } catch (_) {} | |
| _buildFloor(); | |
| _buildPipeline(); | |
| _buildMarker(); | |
| if (!_frameReg) { _stage.onFrame(_onFrame); _frameReg = true; } | |
| _badge = ctx.live.createBadge(); | |
| _polls.push(ctx.live.poll(EP, 5000, _onSpecdecode, { badge: _badge, onState: (m) => { S.state = m.state; _paintOverlay(); } })); | |
| _buildOverlay(); | |
| return { id: ID, started: true }; | |
| } | |
| // ============================================================================= | |
| // builders | |
| // ============================================================================= | |
| function _buildFloor() { | |
| const THREE = _THREE; | |
| const grid = new THREE.GridHelper(40, 40, C_GRID, 0x0f2027); | |
| grid.material.opacity = 0.18; grid.material.transparent = true; grid.position.y = -0.01; | |
| _group.add(grid); | |
| _floor = grid; | |
| } | |
| // Pre-allocate a fixed grid of token meshes: MAX_STEPS steps x MAX_TOKENS tokens | |
| // per step. We toggle visibility / color / position in-place as live data | |
| // arrives (no per-poll geometry churn). | |
| function _buildPipeline() { | |
| const THREE = _THREE; | |
| // pipeline spine: a straight line along X marking the draft->verify axis | |
| { | |
| const pts = [new THREE.Vector3(0, 0, 0), new THREE.Vector3(STEP_LEN * (MAX_STEPS - 1) + 1, 0, 0)]; | |
| const geo = new THREE.BufferGeometry().setFromPoints(pts); | |
| const mat = new THREE.LineBasicMaterial({ color: C_DRAFT, transparent: true, opacity: 0.45 }); | |
| _spine = new THREE.Line(geo, mat); | |
| _group.add(_spine); | |
| } | |
| const tokenGeo = new THREE.OctahedronGeometry(0.20, 0); | |
| for (let s = 0; s < MAX_STEPS; s++) { | |
| const stepTokens = []; | |
| const x = s * STEP_LEN; | |
| for (let t = 0; t < MAX_TOKENS; t++) { | |
| const y = 0.4 + t * TOKEN_GAP; | |
| const mesh = new THREE.Mesh( | |
| tokenGeo, | |
| new THREE.MeshStandardMaterial({ color: C_DRAFT, emissive: C_DRAFT, emissiveIntensity: 0.25, transparent: true, opacity: 0.0 }), | |
| ); | |
| mesh.position.set(x, y, 0); | |
| mesh.visible = false; | |
| _group.add(mesh); | |
| stepTokens.push(mesh); | |
| } | |
| _tokenMesh.push(stepTokens); | |
| // vertical link line from spine up to the step's token column | |
| const linkGeo = new THREE.BufferGeometry().setFromPoints([ | |
| new THREE.Vector3(x, 0, 0), new THREE.Vector3(x, 0.4 + (MAX_TOKENS - 1) * TOKEN_GAP, 0), | |
| ]); | |
| const linkMat = new THREE.LineBasicMaterial({ color: C_GRID, transparent: true, opacity: 0.25 }); | |
| const link = new THREE.Line(linkGeo, linkMat); | |
| link.visible = false; | |
| _group.add(link); | |
| _stepLinks.push(link); | |
| } | |
| } | |
| function _buildMarker() { | |
| const THREE = _THREE; | |
| _marker = new THREE.Mesh( | |
| new THREE.IcosahedronGeometry(0.26, 1), | |
| new THREE.MeshStandardMaterial({ color: C_ACCEPT, emissive: C_ACCEPT, emissiveIntensity: 0.5, wireframe: true, transparent: true, opacity: 0.85 }), | |
| ); | |
| _marker.position.set(0, -0.6, 0); | |
| _group.add(_marker); | |
| } | |
| // ============================================================================= | |
| // live data handler | |
| // ============================================================================= | |
| function _onSpecdecode(j) { | |
| // read honesty label VERBATIM — never upgrade | |
| S.label = (j.label || "MODELED").toUpperCase(); | |
| S.draftLen = typeof j.draft_len === "number" ? j.draft_len : null; | |
| S.trials = typeof j.trials === "number" ? j.trials : null; | |
| S.acceptRate = typeof j.acceptance_rate === "number" ? j.acceptance_rate : null; | |
| S.meanAccept = typeof j.mean_accept_len === "number" ? j.mean_accept_len : null; | |
| S.meanTokens = typeof j.mean_tokens_per_step === "number" ? j.mean_tokens_per_step : null; | |
| S.speedup = typeof j.speedup_factor === "number" ? j.speedup_factor : null; | |
| S.stepLengths = Array.isArray(j.per_step_accept_lengths) ? j.per_step_accept_lengths : null; | |
| S.lossless = typeof j.lossless === "boolean" ? j.lossless : null; | |
| _updatePipeline(); | |
| _paintOverlay(); | |
| } | |
| // ============================================================================= | |
| // geometry updater — drives the token-tree / pipeline from live data | |
| // ============================================================================= | |
| function _updatePipeline() { | |
| const live = S.state === "live"; | |
| const draftLen = live && S.draftLen ? Math.min(S.draftLen, MAX_TOKENS) : 0; | |
| const steps = live && S.stepLengths && S.stepLengths.length | |
| ? S.stepLengths.slice(0, MAX_STEPS) | |
| : []; | |
| for (let s = 0; s < MAX_STEPS; s++) { | |
| const acceptedLen = s < steps.length ? steps[s] : -1; | |
| const showStep = live && s < steps.length && draftLen > 0; | |
| _stepLinks[s].visible = showStep; | |
| for (let t = 0; t < MAX_TOKENS; t++) { | |
| const mesh = _tokenMesh[s][t]; | |
| if (!showStep || t >= draftLen) { | |
| mesh.visible = false; | |
| continue; | |
| } | |
| mesh.visible = true; | |
| const accepted = t < acceptedLen; | |
| const color = accepted ? C_ACCEPT : C_REJECT; | |
| mesh.material.color.setHex(color); | |
| mesh.material.emissive.setHex(color); | |
| mesh.material.emissiveIntensity = accepted ? 0.55 : 0.12; | |
| mesh.material.opacity = accepted ? 0.95 : 0.35; | |
| // rejected tokens (and everything after the first rejection) sit slightly | |
| // recessed in Z to visually read as "pruned" from the accepted path. | |
| mesh.position.z = accepted ? 0 : -0.35; | |
| } | |
| } | |
| // spine + marker degrade to grey when not live | |
| _spine.material.color.setHex(live ? C_DRAFT : C_DIM); | |
| _spine.material.opacity = live ? 0.45 : 0.15; | |
| if (_marker) { | |
| if (live && S.speedup != null) { | |
| _marker.material.color.setHex(C_ACCEPT); | |
| _marker.material.emissive.setHex(C_ACCEPT); | |
| _marker.material.opacity = 0.85; | |
| // marker rides along X proportional to speedup_factor (visual "throughput" cue) | |
| const x = Math.min(STEP_LEN * (MAX_STEPS - 1), (S.speedup || 1) * STEP_LEN * 0.9); | |
| _marker.position.set(x, -0.6, 0); | |
| } else { | |
| _marker.material.color.setHex(C_DIM); | |
| _marker.material.emissive.setHex(C_DIM); | |
| _marker.material.opacity = 0.3; | |
| } | |
| } | |
| } | |
| // ============================================================================= | |
| // per-frame animation | |
| // ============================================================================= | |
| function _onFrame() { | |
| const t = performance.now(); | |
| if (_group) _group.rotation.y = Math.sin(t * 0.00009) * 0.12; | |
| if (_marker) { | |
| _marker.rotation.y += 0.025; | |
| _marker.rotation.x += 0.012; | |
| const pulse = 1.0 + 0.15 * Math.sin(t * 0.004); | |
| _marker.scale.setScalar(pulse); | |
| } | |
| } | |
| // ============================================================================= | |
| // overlay | |
| // ============================================================================= | |
| function _buildOverlay() { | |
| const ctx = _ctx; | |
| _overlay = document.createElement("div"); | |
| Object.assign(_overlay.style, { | |
| position: "absolute", left: "14px", top: "14px", zIndex: "6", | |
| display: "flex", flexDirection: "column", gap: "8px", | |
| maxWidth: "min(94%,440px)", | |
| font: "12px ui-sans-serif,system-ui,Segoe UI,Roboto,Arial", | |
| color: "#eef3f6", | |
| }); | |
| const h = document.createElement("div"); | |
| h.style.cssText = "font:600 13px ui-sans-serif,system-ui;letter-spacing:.4px"; | |
| h.textContent = TITLE; | |
| _overlay.appendChild(h); | |
| const sub = document.createElement("div"); | |
| sub.style.cssText = "color:#9fb1bf;font-size:11px;line-height:1.55"; | |
| sub.innerHTML = | |
| 'A small <b>draft</b> model proposes several tokens per step; the <b>target</b> model ' + | |
| 'verifies them all in one parallel pass, accepting via rejection sampling ' + | |
| '(accept if p<sub>draft</sub> \u2264 p<sub>target</sub>, else w.p. p<sub>target</sub>/p<sub>draft</sub>). ' + | |
| 'Output distribution is <b>lossless</b> \u2014 identical to plain autoregressive decoding, just faster. ' + | |
| 'Honesty label <b>MODELED</b> (deterministic accept/reject simulation; NOT Medusa/EAGLE). 0 runtime CDN.'; | |
| _overlay.appendChild(sub); | |
| const brow = document.createElement("div"); | |
| brow.style.cssText = "display:flex;gap:8px;align-items:center;flex-wrap:wrap"; | |
| if (_badge && _badge.el) brow.appendChild(_badge.el); | |
| _overlay.appendChild(brow); | |
| const card = document.createElement("div"); | |
| card.style.cssText = "background:#0a1117;border:1px solid #1d2a36;border-radius:9px;padding:9px 10px;display:flex;flex-direction:column;gap:6px"; | |
| const chead = document.createElement("div"); | |
| chead.style.cssText = "display:flex;align-items:center;gap:8px;flex-wrap:wrap"; | |
| const dot = document.createElement("span"); | |
| dot.style.cssText = "width:9px;height:9px;border-radius:50%;background:#3af4c8;box-shadow:0 0 7px #3af4c8"; | |
| const nm = document.createElement("b"); | |
| nm.style.cssText = "font-size:12px;color:#3af4c8;letter-spacing:.3px"; | |
| nm.textContent = "self-speculative decoding"; | |
| chead.appendChild(dot); chead.appendChild(nm); | |
| card.appendChild(chead); | |
| const grid = document.createElement("div"); | |
| grid.style.cssText = "display:grid;grid-template-columns:1fr;gap:4px"; | |
| function kpiRow(id, label) { | |
| const r = document.createElement("div"); | |
| r.style.cssText = "display:flex;justify-content:space-between;gap:10px;font-size:11px"; | |
| const l = document.createElement("span"); l.style.cssText = "color:#9fb1bf"; l.textContent = label; | |
| const v = document.createElement("b"); | |
| v.id = id; | |
| v.style.cssText = "font-variant-numeric:tabular-nums;color:#eef3f6;text-align:right;max-width:58%"; | |
| v.textContent = "\u2014"; | |
| _el[id] = v; | |
| r.appendChild(l); r.appendChild(v); return r; | |
| } | |
| grid.appendChild(kpiRow("sd-draftlen", "draft_len (\u03b3 proposed/step)")); | |
| grid.appendChild(kpiRow("sd-trials", "trials simulated")); | |
| grid.appendChild(kpiRow("sd-accrate", "acceptance_rate \u2014 MODELED")); | |
| grid.appendChild(kpiRow("sd-meanacc", "mean accepted tokens/step")); | |
| grid.appendChild(kpiRow("sd-speedup", "speedup_factor \u2014 MODELED")); | |
| grid.appendChild(kpiRow("sd-lossless", "lossless?")); | |
| grid.appendChild(kpiRow("sd-label", "honesty label")); | |
| card.appendChild(grid); | |
| const fn = document.createElement("div"); | |
| fn.style.cssText = "font-size:9.5px;color:#6b7a86;line-height:1.5"; | |
| fn.textContent = "Leviathan et al. arXiv:2211.17192 (speculative decoding) \u00b7 Medusa github.com/FasterDecoding/Medusa \u00b7 EAGLE arXiv:2401.15077. MODELED \u00b7 not claimed-as."; | |
| card.appendChild(fn); | |
| _overlay.appendChild(card); | |
| const pl = document.createElement("button"); | |
| pl.textContent = "\u25d1 what this means"; | |
| pl.title = "Toggle plain-language explanation for investors & consumers."; | |
| pl.style.cssText = "font:11px ui-monospace,monospace;padding:5px 11px;border-radius:7px;border:1px solid #3af4c8;background:#08140f;color:#3af4c8;cursor:pointer;width:fit-content"; | |
| pl.addEventListener("click", () => { | |
| _plain = !_plain; | |
| pl.style.background = _plain ? "#0f2a20" : "#08140f"; | |
| _applyPlain(); | |
| }); | |
| _overlay.appendChild(pl); | |
| const pd = document.createElement("div"); | |
| pd.id = "sd-plain"; | |
| pd.style.cssText = "font-size:10.5px;color:#c9d6df;line-height:1.55;border:1px dashed #26333f;border-radius:7px;padding:7px 9px;display:none"; | |
| _el["plain"] = pd; | |
| _overlay.appendChild(pd); | |
| (ctx.container || document.body).appendChild(_overlay); | |
| _paintOverlay(); | |
| } | |
| function _applyPlain() { | |
| const pd = _el["plain"]; | |
| if (!pd) return; | |
| pd.style.display = _plain ? "block" : "none"; | |
| if (!_plain) return; | |
| const g = S.draftLen != null ? String(S.draftLen) : "loading\u2026"; | |
| const accPct = S.acceptRate != null ? (S.acceptRate * 100).toFixed(1) + "%" : "loading\u2026"; | |
| const speedup = S.speedup != null ? S.speedup.toFixed(2) + "\u00d7" : "loading\u2026"; | |
| pd.innerHTML = | |
| "<b>What this means:</b> Instead of generating one word at a time, a small, cheap " + | |
| "\u201cdraft\u201d model quickly guesses <b>" + g + " words ahead</b>. The big \u201ctarget\u201d " + | |
| "model then checks all of those guesses <i>at once</i> and keeps every guess it agrees with " + | |
| "(here, about <b>" + accPct + "</b> of guesses survive). Because checking a whole batch of " + | |
| "guesses costs about the same as generating one word normally, this yields roughly a <b>" + | |
| speedup + " speedup</b> in tokens produced per expensive model pass \u2014 with " + | |
| "<i>zero change</i> to what the model actually outputs (mathematically lossless, per " + | |
| "Leviathan et al. 2023). Plain: same answer, produced faster. This view is a <b>MODELED</b> " + | |
| "closed-form simulation of the accept/reject math, not a run of Medusa or EAGLE."; | |
| } | |
| function _tok(s) { | |
| if (s === "live") return null; | |
| if (s === "missing") return "NO-LIVE-DATA"; | |
| if (s === "degraded") return "DEGRADED"; | |
| if (s === "error") return "OFFLINE"; | |
| return "\u2026"; | |
| } | |
| function fx(v, d) { return typeof v === "number" ? v.toFixed(d) : "\u2014"; } | |
| function pct(v, d) { return typeof v === "number" ? (v * 100).toFixed(d) + "%" : "\u2014"; } | |
| function _set(id, v) { if (_el[id]) _el[id].textContent = v; } | |
| function _paintOverlay() { | |
| const t = _tok(S.state); | |
| _set("sd-draftlen", t || (S.draftLen != null ? String(S.draftLen) : "\u2014")); | |
| _set("sd-trials", t || (S.trials != null ? String(S.trials) : "\u2014")); | |
| _set("sd-accrate", t || pct(S.acceptRate, 2)); | |
| _set("sd-meanacc", t || fx(S.meanAccept, 3)); | |
| _set("sd-speedup", t || (S.speedup != null ? S.speedup.toFixed(3) + "\u00d7" : "\u2014")); | |
| _set("sd-lossless", t || (S.lossless === true ? "yes (rejection-sampling identity)" : S.lossless === false ? "no" : "\u2014")); | |
| // honesty label verbatim — never upgraded | |
| _set("sd-label", t || (S.label || "MODELED")); | |
| if (_plain) _applyPlain(); | |
| } | |
| // ============================================================================= | |
| // unmount — clean up everything; must not affect other organs | |
| // ============================================================================= | |
| export function unmount() { | |
| _polls.forEach((p) => { try { p.stop(); } catch (_) {} }); _polls = []; | |
| try { if (_overlay && _overlay.parentNode) _overlay.parentNode.removeChild(_overlay); } catch (_) {} | |
| try { | |
| if (_group && _stage) { | |
| _group.traverse((o) => { | |
| if (o.geometry && o.geometry.dispose) o.geometry.dispose(); | |
| if (o.material) { | |
| const ms = Array.isArray(o.material) ? o.material : [o.material]; | |
| ms.forEach((m) => { if (m.dispose) m.dispose(); }); | |
| } | |
| }); | |
| _stage.scene.remove(_group); | |
| } | |
| } catch (_) {} | |
| _group = _overlay = null; | |
| _floor = null; _spine = null; _tokenMesh = []; _stepLinks = []; _marker = null; | |
| _el = {}; _badge = null; _plain = false; _frameReg = false; | |
| _stage = _THREE = _ctx = null; | |
| S.label = S.draftLen = S.trials = S.acceptRate = null; | |
| S.meanAccept = S.meanTokens = S.speedup = S.stepLengths = S.lossless = null; | |
| S.state = "init"; | |
| } | |
| export default { id: ID, title: TITLE, endpoints: [EP], mount, unmount }; | |