Text Generation
Transformers
Chinese
English
character-level
chinese
gru
mamba
Mixture of Experts
titans
Instructions to use masterofthestars/tianxuan with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use masterofthestars/tianxuan with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="masterofthestars/tianxuan")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("masterofthestars/tianxuan", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use masterofthestars/tianxuan with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "masterofthestars/tianxuan" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "masterofthestars/tianxuan", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/masterofthestars/tianxuan
- SGLang
How to use masterofthestars/tianxuan with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "masterofthestars/tianxuan" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "masterofthestars/tianxuan", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "masterofthestars/tianxuan" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "masterofthestars/tianxuan", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use masterofthestars/tianxuan with Docker Model Runner:
docker model run hf.co/masterofthestars/tianxuan
Tianxuan: full project + data (checkpoints, world_model, data without raw) with correct directory structure
db7bc06 Download src/ml/math_engine.ts from masterofthestars/tianxuan: direct link, hf CLI and curl.
- Browser
- Download file 15 kB
-
https://huggingface.co/masterofthestars/tianxuan/resolve/main/src/ml/math_engine.ts
- Command line
-
hf download hf://masterofthestars/tianxuan/src/ml/math_engine.ts
-
curl -L -o math_engine.ts https://huggingface.co/masterofthestars/tianxuan/resolve/main/src/ml/math_engine.ts
15 kB
| /** | |
| * SymbolicMathEngine — 符号数学推理引擎 | |
| * | |
| * 不追求数值近似, 追求概念理解: | |
| * - 常量: 定义 + 性质 + 已知展开式 (pi, e, phi, sqrt2 等) | |
| * - 定理: 陈述 + 适用条件 + 推广/逆命题 | |
| * - 输出不是单一数字, 而是"理解包" | |
| */ | |
| export interface ConstantInfo { | |
| name: string; | |
| symbol: string; | |
| definition: string; | |
| approximation: string; | |
| properties: string[]; | |
| appearances: string[]; | |
| related: string[]; | |
| } | |
| export interface TheoremInfo { | |
| name: string; | |
| symbol: string; | |
| statement: string; | |
| conditions: string[]; | |
| proof_idea: string; | |
| generalizations: string[]; | |
| counterexamples: string[]; | |
| applications: string[]; | |
| related: string[]; | |
| } | |
| export interface CalculationResult { | |
| question: string; | |
| exactAnswer: string; | |
| approximate?: string; | |
| derivation: string[]; | |
| explanation: string; | |
| relatedInfo?: { type: "constant" | "theorem"; key: string }; | |
| } | |
| const CONSTANT_DB: Record<string, ConstantInfo> = { | |
| pi: { | |
| name: "圆周率", | |
| symbol: "π", | |
| definition: "圆的周长与直径之比, pi = C/d = 2*pi*r/r = 2A/r^2, 其中 A 为圆面积", | |
| approximation: "3.14159265358979323846264338327950288419716939937510...", | |
| properties: [ | |
| "无理数: 不能表示为两个整数之比 (Lambert, 1768)", | |
| "超越数: 不是任何整系数多项式方程的根 (Lindemann, 1882)", | |
| "因此圆方问题尺规作图不可能", | |
| "欧拉恒等式: e^(i*pi) + 1 = 0, 连接五个基本常数", | |
| "出现在概率中: 布丰投针问题 P(交叉) = 2l/(pi*d)", | |
| "傅里叶变换: integral_{-inf}^{inf} e^(-ix*xi) dxi = 2*pi*delta(x)", | |
| ], | |
| appearances: [ | |
| "圆面积 S = pi*r^2, 圆周长 C = 2*pi*r", | |
| "球体积 V = (4/3)*pi*r^3, 球表面积 S = 4*pi*r^2", | |
| "正态分布密度: f(x) = (1/(sigma*sqrt(2*pi))) * exp(-(x-mu)^2/(2*sigma^2))", | |
| "欧拉公式: e^(i*theta) = cos(theta) + i*sin(theta)", | |
| "黎曼 zeta 函数: zeta(2) = pi^2/6 (巴塞尔问题)", | |
| ], | |
| related: ["e", "i", "phi", "欧拉恒等式", "傅里叶分析", "复分析"], | |
| }, | |
| e: { | |
| name: "自然常数", | |
| symbol: "e", | |
| definition: "e = lim(n->inf)(1+1/n)^n = sum(n=0..inf) 1/n! ≈ 2.71828..., 是指数函数 d/dx[e^x]=e^x 的底", | |
| approximation: "2.71828182845904523536028747135266249775724709369995...", | |
| properties: [ | |
| "无理数 (Euler, 1737)", | |
| "超越数 (Hermite, 1873)", | |
| "e^x 是唯一满足 f'=f, f(0)=1 的函数", | |
| "ln(e) = 1, log_e 是以 e 为底的对数", | |
| "e^(i*pi) = -1 (欧拉恒等式)", | |
| ], | |
| appearances: [ | |
| "指数增长/衰减: N(t) = N_0 * e^(rt)", | |
| "复利极限: lim(n->inf)(1+r/n)^(nt) = e^(rt)", | |
| "正态分布, 泊松分布, 指数分布", | |
| "双曲函数: cosh(x) = (e^x + e^(-x))/2", | |
| ], | |
| related: ["pi", "i", "欧拉恒等式", "微积分", "微分方程"], | |
| }, | |
| phi: { | |
| name: "黄金比例", | |
| symbol: "phi", | |
| definition: "phi = (1+sqrt(5))/2, 满足 phi^2 = phi+1, 即 phi/1 = 1/(phi-1)", | |
| approximation: "1.61803398874989484820458683436563811772030917980576...", | |
| properties: [ | |
| "无理数 (古希腊已知)", | |
| "phi = 1 + 1/phi, 连分数 [1;1,1,1,...] 无限循环", | |
| "phi^2 = phi+1, phi^3 = 2*phi+1", | |
| "1(phi) = phi - 1 ≈ 0.618", | |
| "与斐波那契数列关系: F_n / F_(n-1) -> phi", | |
| ], | |
| appearances: [ | |
| "正五边形对角线/边长 = phi", | |
| "斐波那契螺旋, 向日葵种子排列", | |
| "黄金矩形, 美学中的经典比例", | |
| ], | |
| related: ["sqrt5", "斐波那契数列", "连分数", "正五边形"], | |
| }, | |
| sqrt2: { | |
| name: "根号2", | |
| symbol: "sqrt(2)", | |
| definition: "sqrt(2) 是满足 x^2 = 2 的正实数, 即单位正方形对角线长度", | |
| approximation: "1.41421356237309504880168872420969807856967187537694...", | |
| properties: [ | |
| "无理数 (毕达哥拉斯学派, 希帕索斯发现)", | |
| "最早被证明的无理数之一", | |
| "sqrt(2) 不是任何整系数二次方程的有理根", | |
| ], | |
| appearances: [ | |
| "勾股定理特例: 直角边为1的等腰直角三角形斜边", | |
| "单位正方形对角线长", | |
| ], | |
| related: ["勾股定理", "无理数", "毕达哥拉斯"], | |
| }, | |
| }; | |
| const THEOREM_DB: Record<string, TheoremInfo> = { | |
| pythagoras: { | |
| name: "勾股定理", | |
| symbol: "a^2 + b^2 = c^2", | |
| statement: "在欧几里得平面中, 任意直角三角形的两条直角边的平方和等于斜边的平方", | |
| conditions: ["三角形是直角三角形", "处于欧几里得平面 (平坦空间)", "二维或更高维欧氏空间"], | |
| proof_idea: "经典证法: 相似三角形法、面积拼接法 (赵爽弦图)、向量法 (|a+b|^2=|a|^2+|b|^2 当 a perp b)。现代观点: 勾股定理等价于内积空间的范数平行四边形恒等式", | |
| generalizations: [ | |
| "余弦定理: c^2 = a^2+b^2-2ab*cos(C) (一般三角形的推广)", | |
| "勾股定理在黎曼流形中不再成立 (测地线偏离)", | |
| "高维: 勾股定理对正交向量组的范数仍然成立", | |
| ], | |
| counterexamples: [ | |
| "球面三角形: 三面角之和 > 180度, 勾股定理不适用 (非欧几何)", | |
| "双曲几何: 直角三角形斜边 > sqrt(a^2+b^2)", | |
| "曼哈顿距离: d = |x1-x2| + |y1-y2|, 勾股定理形式不成立", | |
| ], | |
| applications: [ | |
| "计算两点间欧氏距离: d = sqrt((x2-x1)^2+(y2-y1)^2)", | |
| "向量内积: a.b = |a||b|cos(theta), 勾股定理是 theta=90度的特例", | |
| "Parseval 恒等式是勾股定理在无穷维的推广", | |
| ], | |
| related: ["余弦定理", "欧几里得几何", "黎曼几何", "内积空间"], | |
| }, | |
| euler_identity: { | |
| name: "欧拉恒等式", | |
| symbol: "e^(i*pi) + 1 = 0", | |
| statement: "自然常数 e, 虚数单位 i, 圆周率 pi, 加法单位元 0, 乘法单位元 1 通过基本运算联系在一起", | |
| conditions: ["复指数函数 e^z 的解析延拓成立", "复平面上极坐标表示有效"], | |
| proof_idea: "由欧拉公式 e^(i*theta) = cos(theta) + i*sin(theta), 代入 theta = pi: e^(i*pi) = cos(pi) + i*sin(pi) = -1 + 0*i = -1, 移项即得", | |
| generalizations: ["欧拉公式: e^(i*theta) = cos(theta) + i*sin(theta)", "复对数: log(z) = ln|z| + i(arg(z) + 2*k*pi), 多值性"], | |
| counterexamples: [], | |
| applications: [ | |
| "交流电路分析: 阻抗的复数表示", | |
| "傅里叶变换的基础", | |
| "量子力学: 波函数的相位因子 e^(i*theta)", | |
| ], | |
| related: ["欧拉公式", "复分析", "傅里叶分析", "pi", "e", "i"], | |
| }, | |
| schrodinger: { | |
| name: "薛定谔方程", | |
| symbol: "i*hbar*dPsi/dt = H*Psi", | |
| statement: "量子力学的基本运动方程, 描述量子态 (波函数) 随时间的演化", | |
| conditions: ["非相对论性量子力学 (粒子速度 << c)", "势场不显含时间时可用分离变量法", "单粒子近似"], | |
| proof_idea: "薛定谔方程是量子力学的公设之一, 不能从更基本的原理推导。思路: 从德布罗意物质波关系 E=p^2/(2m)+V 和平面波 Psi=A*e^(i(kx-omega*t)) 出发, 替换 E->i*hbar*d/dt, p->-i*hbar*grad, 得到含时薛定谔方程", | |
| generalizations: [ | |
| "含势场的含时薛定谔方程: i*hbar*dPsi/dt = [-hbar^2/(2m)*laplacian + V(r,t)]*Psi", | |
| "定态薛定谔方程: H*psi = E*psi (分离变量, V 不依赖 t)", | |
| "狄拉克方程: 相对论性自旋 1/2 粒子的推广", | |
| ], | |
| counterexamples: [ | |
| "光速粒子 (光子): 需用量子电动力学", | |
| "强引力场: 需结合广义相对论", | |
| "多体相互作用: 严格求解困难, 需近似方法", | |
| ], | |
| applications: [ | |
| "氢原子能级计算 (玻尔模型的量子力学解释)", | |
| "谐振子能级 E_n = (n+1/2)*hbar*omega", | |
| "势垒隧穿效应 (扫描隧道显微镜原理)", | |
| ], | |
| related: ["海森堡矩阵力学", "费曼路径积分", "算子理论", "希尔伯特空间"], | |
| }, | |
| }; | |
| export class SymbolicMathEngine { | |
| answer(question: string): CalculationResult { | |
| const q = question.trim().toLowerCase(); | |
| for (const [key, info] of Object.entries(CONSTANT_DB)) { | |
| if (q.includes(info.symbol) || q.includes(info.name) || q.includes(key)) { | |
| return this._buildResult(question, info); | |
| } | |
| } | |
| for (const [key, info] of Object.entries(THEOREM_DB)) { | |
| if (q.includes(info.name) || q.includes(key) || q.includes(info.symbol)) { | |
| return this._buildTheoremResult(question, info); | |
| } | |
| } | |
| const numMatch = q.match(/(\d+\.?\d*)\s*[\*\/\+\-\^]/); | |
| if (numMatch) { | |
| return this._handleExpression(q); | |
| } | |
| if (q.includes("无理") || q.includes("超越") || q.includes("有理")) { | |
| return this._answerNumberTypeQuestion(q); | |
| } | |
| return this._buildDefaultResult(question); | |
| } | |
| private _buildResult(q: string, info: ConstantInfo): CalculationResult { | |
| return { | |
| question: q, | |
| exactAnswer: info.symbol, | |
| approximate: info.approximation + "...(无限不循环, 无法完整写出)", | |
| derivation: [ | |
| "定义: " + info.definition, | |
| ...info.properties.map((p, i) => (i + 1) + ". " + p), | |
| "数值展开: " + info.approximation.slice(0, 60) + "...(无限位数, 无周期)", | |
| ], | |
| explanation: "你问「" + q + "」—— 这不是一个能'算出有限位就结束'的问题。" + | |
| info.symbol + " 是一个无限不循环的数, 任何有限近似都只是近似。" + | |
| "它之所以重要, 不是因为它的数值, 而是因为它出现在" + info.appearances.length + "个不同领域的基础公式中:" + | |
| info.appearances.slice(0, 3).map((a, i) => (i + 1) + ". " + a).join("\n"), | |
| relatedInfo: { type: "constant", key: Object.entries(CONSTANT_DB).find(([, v]) => v.symbol === info.symbol)?.[0] ?? "" }, | |
| }; | |
| } | |
| private _buildTheoremResult(q: string, info: TheoremInfo): CalculationResult { | |
| return { | |
| question: q, | |
| exactAnswer: info.symbol, | |
| derivation: [ | |
| "定理: " + info.statement, | |
| "适用条件: " + info.conditions.join(", "), | |
| "证明思路: " + info.proof_idea, | |
| ], | |
| explanation: "「" + info.name + "」不是一个公式可以简单套用的工具, 而是一个有适用边界的数学结构。" + | |
| "它的核心不是\"" + info.symbol + "\"这行字, 而是理解:" + | |
| "\n -> 为什么在欧氏空间成立 (内积结构的必然结果)" + | |
| "\n -> 为什么在非欧几何失效 (空间弯曲了)" + | |
| "\n -> 它的推广是什么 (" + info.generalizations.slice(0, 2).join("; ") + ")" + | |
| (info.counterexamples.length > 0 ? "\n\n反例: " + info.counterexamples.slice(0, 2).join("; ") : ""), | |
| relatedInfo: { type: "theorem", key: Object.entries(THEOREM_DB).find(([, v]) => v.name === info.name)?.[0] ?? "" }, | |
| }; | |
| } | |
| private _handleExpression(q: string): CalculationResult { | |
| try { | |
| // 安全数学表达式求值: 只用数字、运算符、括号、空格, 不做代码注入 | |
| const cleaned = q.replace(/[^0-9+\-*/^.()\s*pi]/g, "").trim(); | |
| if (!cleaned) return this._buildDefaultResult(q); | |
| // 用递归下降解析器求值, 避免 Function/eval | |
| const result = this._safeEval(cleaned); | |
| if (typeof result === "number" && isFinite(result)) { | |
| return { | |
| question: q, | |
| exactAnswer: q, | |
| approximate: result.toFixed(10), | |
| derivation: ["代入数值计算: " + q + " = " + result.toFixed(10)], | |
| explanation: "数值结果是 " + result.toFixed(10) + ", 但注意: 如果表达式中包含 pi, 这个结果是近似的。" + | |
| "pi 本身是无限不循环的, 任何有限十进制表示都只是近似值。", | |
| }; | |
| } | |
| } catch (_) { /* fall through */ } | |
| return this._buildDefaultResult(q); | |
| } | |
| /** 递归下降安全求值: 仅支持 + - * / ^ () 和数字/pi */ | |
| private _safeEval(expr: string): number { | |
| let pos = 0; | |
| const source = expr.replace(/\s+/g, ""); | |
| const peek = (): string => source[pos] ?? ""; | |
| const consume = (): string => source[pos++]; | |
| const parseNumber = (): number => { | |
| let numStr = ""; | |
| while (pos < source.length && (/[0-9.]/.test(source[pos]))) numStr += consume(); | |
| if (!numStr) throw new Error("unexpected char at " + pos); | |
| const v = parseFloat(numStr); | |
| if (isNaN(v)) throw new Error("invalid number: " + numStr); | |
| return v; | |
| }; | |
| const parseConstant = (): number => { | |
| if (source.slice(pos, pos + 2) === "pi") { pos += 2; return Math.PI; } | |
| return NaN; | |
| }; | |
| const parseTerm = (): number => { | |
| let left: number; | |
| if (peek() === "(") { | |
| consume(); // '(' | |
| left = parseExpr(); | |
| if (consume() !== ")") throw new Error("missing )"); | |
| } else if (/^[0-9.]/.test(peek())) { | |
| left = parseNumber(); | |
| } else { | |
| const c = parseConstant(); | |
| if (isNaN(c)) throw new Error("unexpected: " + peek()); | |
| left = c; | |
| } | |
| while (pos < source.length && (source[pos] === "*" || source[pos] === "/" || source[pos] === "^")) { | |
| const op = consume(); | |
| const right = parseTerm(); | |
| if (op === "*") left *= right; | |
| else if (op === "/") { if (right === 0) throw new Error("div by zero"); left /= right; } | |
| else if (op === "^") left = Math.pow(left, right); | |
| } | |
| return left; | |
| }; | |
| const parseExpr = (): number => { | |
| let left = parseTerm(); | |
| while (pos < source.length && (source[pos] === "+" || source[pos] === "-")) { | |
| const op = consume(); | |
| const right = parseTerm(); | |
| left = op === "+" ? left + right : left - right; | |
| } | |
| return left; | |
| }; | |
| const result = parseExpr(); | |
| if (pos !== source.length) throw new Error("trailing chars at " + pos); | |
| return result; | |
| } | |
| private _answerNumberTypeQuestion(q: string): CalculationResult { | |
| if (q.includes("pi") || q.includes("圆周率")) return this._buildResult(q, CONSTANT_DB.pi); | |
| if (q.includes("e") && (q.includes("自然") || q.includes("常数"))) return this._buildResult(q, CONSTANT_DB.e); | |
| return this._buildDefaultResult(q); | |
| } | |
| private _buildDefaultResult(q: string): CalculationResult { | |
| return { | |
| question: q, | |
| exactAnswer: "未知", | |
| derivation: ["未能在知识库中找到匹配的数学概念或定理"], | |
| explanation: "这个问题目前没有匹配到的精确数学知识。" + | |
| "如果你有具体的数学问题 (常量/定理/方程), 可以用更明确的术语提问," + | |
| "例如: " + String.fromCharCode(8216) + "pi 等于多少" + String.fromCharCode(8217) + ", " + String.fromCharCode(8216) + "勾股定理是什么" + String.fromCharCode(8217) + ", " + String.fromCharCode(8216) + "薛定谔方程怎么解" + String.fromCharCode(8217) + ".", | |
| }; | |
| } | |
| } | |
| /** 别名,供 reasoner.ts / symbolic_engine.ts 使用 */ | |
| export const MathEngine = SymbolicMathEngine; | |