总览 · ← 返回运行 20261002-202908-search-t1-scr-C

节点 n59

改了什么

方法说明?节点程序自带的 METHOD.md：这个程序做了什么、为什么。

来自 programs.git e5de4f3044:solution/METHOD.md

调研员的计划

代码改动?这个节点的程序和父节点程序的逐行差别：绿色是新增，红色是删除。

对比：这个提交的上一版（种子程序：相对空仓库）。改动的文件：solution/EXECUTION.json +1 −0、solution/METHOD.md +16 −0、solution/README.md +4 −0、solution/g37_common.py +101 −0、solution/run.py +156 −0

diff --git a/solution/EXECUTION.json b/solution/EXECUTION.jsonnew file mode 100644index 0000000..9d5125c--- /dev/null+++ b/solution/EXECUTION.json@@ -0,0 +1 @@+{"gpu": false}diff --git a/solution/METHOD.md b/solution/METHOD.mdnew file mode 100644index 0000000..33a0cb8--- /dev/null+++ b/solution/METHOD.md@@ -0,0 +1,16 @@+## 改了什么+将 ot_moscot（Waddington-OT 耦合 + 重心位移外推）替换为 stochastic_bridge（Schrödinger Bridge / [SF]²M）：+1. 在联合 PCA 空间用 torchcfm 的 SchrodingerBridgeConditionalFlowMatcher(sigma=0.3) 训练时间依赖速度场（3层 MLP，128 hidden，400 epochs，Adam lr=1e-3）；+2. 配对策略：对 last 阶段每个细胞在 PCA 空间找 prev 的最近邻（kNN pairing），训练时按此配对采样 mini-batch；+3. 推理：对输出细胞在 t=0.95 处评估学到的漂移（避免 t=1 处桥的奇异性），取 STEP_FRACTION=0.6 × dt_out/dt_in 的步长；+4. 随机性：添加 NOISE_SCALE×σ×√(dt_out/dt_in) 的高斯噪声（σ=0.3），体现桥的扩散不确定性；+5. 解码：PCA 步长通过 components×sd 解码到基因空间（仅 HVG），addnz 只改非零位，夹到 ≥0；+6. 机制关闭对照：环境变量 SB_MECHANISM_OFF=1 时 σ=0（退化为确定性 CFM，无桥噪声），提交时保持打开；+7. 单输入阶段退路不变：生长加权复制。+与父节点的核心区别：漂移由神经网络给出（状态依赖），而非重心投影（全局平均）；加入受控噪声保留群体扩散。+## 用到的知识与出处+- Tong A. et al., Simulation-free Schrödinger bridges ([SF]²M), 2024, arXiv:2307.03672（桥匹配器公式、熵正则）+- torchcfm 1.0.7（MIT），SchrodingerBridgeConditionalFlowMatcher API（知识卡 k034）+- Lipman Y. et al., Flow Matching for Generative Modeling, ICLR 2023, arXiv:2210.02747（条件流匹配训练目标）+- 生长率先验：Schiebinger et al. Cell 2019（WOT birth-death logistic，同父节点）+- PCA 嵌入与 addnz 解码策略：继承自父节点 g37_common.pydiff --git a/solution/README.md b/solution/README.mdnew file mode 100644index 0000000..dc1cbaf--- /dev/null+++ b/solution/README.md@@ -0,0 +1,4 @@+# ot_moscot++Waddington-OT / moscot：两个最新输入阶段在联合 PCA 上做非平衡熵 OT 耦合（WOT 增殖/凋亡先验生长），最新阶段按 g^Δt 重抽样，每个细胞沿“kNN 平滑位置 − 耦合祖先均值”再走一个等比例步长，只加在非零基因上，夹到 ≥0。+来自 G37 候选 modeling/candidates/T1/ot_moscot；addnz 解码是在 X3 上选的（详见 METHOD.md）。纯 CPU，final 约 70 s、6.7 GB。diff --git a/solution/g37_common.py b/solution/g37_common.pynew file mode 100644index 0000000..8c991b3--- /dev/null+++ b/solution/g37_common.py@@ -0,0 +1,101 @@+"""Helpers of the ot_moscot seed (copy of modeling/candidates/T1/ot_moscot/g37_common.py, G37; stage_pair reads every+input stage of the view).++- ``stage_pair``: the two latest input stages (official, or external in proxy2 / test views) and the gene mask on+  which both are really measured (external stages lack part of the panel; filled columns must not drive dynamics);+- ``embed``: HVG -> z-score (clip 10) -> PCA, fitted on the input stages only (scanpy-style preprocessing);+- ``knn_mean``: per-cell kNN average in the embedding (smooths single-cell noise before taking displacements);+- ``growth_rates``: Waddington-OT / moscot prior growth from proliferation and apoptosis gene scores.+"""++from __future__ import annotations++import numpy as np+from scipy import sparse++from src.task1_temporal.view_io import covered_mask, inputs_by_time, read_stage+++def stage_pair(view, manifest: dict, genes: list[str]):+    """(prev, last, mask, dt_in) for the two latest inputs; prev None with a single input stage."""+    stages = inputs_by_time(manifest, include_external=True)  # every input stage, whatever its source+    last_e = stages[-1]+    last = read_stage(view, last_e, genes)+    mask = covered_mask(view, last_e, genes)+    if len(stages) < 2:+        return None, last, mask, None, last_e+    prev_e = stages[-2]+    prev = read_stage(view, prev_e, genes)+    mask &= covered_mask(view, prev_e, genes)+    return prev, last, mask, float(last_e["time"] - prev_e["time"]), last_e+++def _gene_moments(X: sparse.csr_matrix):+    n = X.shape[0]+    mean = np.asarray(X.mean(axis=0)).ravel()+    sq = np.asarray(X.multiply(X).sum(axis=0)).ravel() / n+    return mean, np.maximum(sq - mean ** 2, 0.0)+++def embed(mats: list, mask: np.ndarray, n_hvg: int, n_pcs: int, seed: int):+    """PCA coordinates of each matrix in ``mats`` (same order), fitted on all of them together.++    Returns (list of Z, info) with info = {hvg, mean, std, components} so displacements can be decoded."""+    from sklearn.decomposition import PCA++    allX = sparse.vstack(mats).tocsr()+    mean, var = _gene_moments(allX)+    var = np.where(mask, var, -1.0)+    hvg = np.sort(np.argsort(var)[::-1][:n_hvg])+    sub = allX[:, hvg].toarray().astype(np.float32)+    mu, sd = mean[hvg].astype(np.float32), np.sqrt(var[hvg]).astype(np.float32)+    sd[sd == 0] = 1.0+    sub -= mu+    sub /= sd+    np.clip(sub, -10, 10, out=sub)+    pca = PCA(n_components=n_pcs, svd_solver="randomized", random_state=seed)+    Z = pca.fit_transform(sub).astype(np.float32)+    out, start = [], 0+    for M in mats:+        out.append(Z[start:start + M.shape[0]])+        start += M.shape[0]+    return out, {"hvg": hvg, "mu": mu, "sd": sd, "components": pca.components_.astype(np.float32)}+++def knn_mean(Z: np.ndarray, X: sparse.csr_matrix, rows: np.ndarray, k: int) -> np.ndarray:+    """Dense (len(rows), n_genes): mean expression of the k nearest cells (in Z, the cell itself included)."""+    from sklearn.neighbors import NearestNeighbors++    nn = NearestNeighbors(n_neighbors=k).fit(Z)+    idx = nn.kneighbors(Z[rows], return_distance=False)+    W = sparse.csr_matrix((np.full(idx.size, 1.0 / k, dtype=np.float32), idx.ravel(),+                           np.arange(0, idx.size + 1, k)), shape=(len(rows), Z.shape[0]))+    return np.asarray((W @ X).todense(), dtype=np.float32)+++# Waddington-OT prior growth (Schiebinger et al. 2019, Cell; as implemented in moscot.base.problems.birth_death):+# birth = generalised logistic of the proliferation score, death = of the apoptosis score, g = exp(birth - death) / day.+def _gen_logistic(p, sup, inf, center, width):+    return inf + (sup - inf) / (1 + np.exp(-(p - center) / width))+++def growth_rates(adata_list: list, genes: list[str], mask: np.ndarray, seed: int) -> list[np.ndarray]:+    """Per-day growth rate of every cell of each AnnData (scores computed on all of them jointly)."""+    import anndata as ad+    import scanpy as sc+    from moscot.utils.data import apoptosis_markers, proliferation_markers++    ok = set(np.asarray(genes)[mask])+    joint = ad.concat(adata_list, join="inner")+    prolif = [g for g in proliferation_markers("mouse") if g in ok]+    apopt = [g for g in apoptosis_markers("mouse") if g in ok]+    sc.tl.score_genes(joint, prolif, score_name="proliferation", random_state=seed)+    sc.tl.score_genes(joint, apopt, score_name="apoptosis", random_state=seed)+    birth = _gen_logistic(joint.obs["proliferation"].to_numpy(float), 1.7, 0.3, 0.25, 0.5)+    death = _gen_logistic(joint.obs["apoptosis"].to_numpy(float), 1.7, 0.3, 0.1, 0.2)+    g = np.exp(birth - death)+    out, start = [], 0+    for a in adata_list:+        out.append(g[start:start + a.n_obs])+        start += a.n_obs+    return outdiff --git a/solution/run.py b/solution/run.pynew file mode 100644index 0000000..6d9b105--- /dev/null+++ b/solution/run.py@@ -0,0 +1,156 @@+#!/usr/bin/env python3+"""stochastic_bridge: Schrodinger Bridge ([SF]2M) flow matching in PCA, drift extrapolation + controlled noise.++Two input stages:+  1. Joint PCA of both stages (HVG, z-score, 30 PCs; fitted on inputs only);+  2. kNN pairing (last -> prev in PCA space);+  3. Train SchrodingerBridgeConditionalFlowMatcher (torchcfm) velocity field on paired samples;+  4. Output cells = latest-stage cells resampled by growth^dt_out (WOT birth-death prior);+  5. Evaluate drift at t=0.95 in PCA for output cells, take extrapolation step, add controlled noise;+  6. Decode step to gene space (HVGs only), apply addnz, clip at 0.++One input stage: growth-weighted copy (fallback).++Mechanism-off control: SB_MECHANISM_OFF=1 sets sigma=0 (deterministic CFM, no bridge noise).+"""++from __future__ import annotations++import argparse+import os+import sys+from pathlib import Path++import numpy as np++sys.path.insert(0, str(Path(__file__).resolve().parent))+from g37_common import embed, growth_rates, stage_pair  # noqa: E402++from src.task1_temporal.view_io import load_manifest, panel_genes, target_n_cells, write_prediction  # noqa: E402++N_HVG = 2000+N_PCS = 30+SIGMA = 0.3+STEP_FRACTION = 0.6+NOISE_SCALE = 0.4+T_EVAL = 0.95+N_EPOCHS = 400+HIDDEN = 128+LR = 1e-3+BATCH_SIZE = 512+N_THREADS = 8+MECHANISM_OFF = os.environ.get("SB_MECHANISM_OFF", "0") == "1"+++def weighted_rows(w: np.ndarray, n: int, rng: np.random.Generator) -> np.ndarray:+    p = w / w.sum()+    return np.sort(rng.choice(len(w), size=n, replace=n > len(w), p=p))+++def main() -> None:+    parser = argparse.ArgumentParser()+    parser.add_argument("--data", required=True)+    parser.add_argument("--out", required=True)+    parser.add_argument("--seed", type=int, default=0)+    args = parser.parse_args()++    manifest = load_manifest(args.data)+    genes = panel_genes(args.data, manifest)+    prev, last, mask, dt_in, last_e = stage_pair(args.data, manifest, genes)+    dt_out = float(manifest["target"]["time"] - last_e["time"])+    rng = np.random.default_rng(args.seed)+    n = target_n_cells(manifest, last.n_obs)++    if prev is None:+        (g_last,) = growth_rates([last], genes, mask, args.seed)+        rows = weighted_rows(g_last ** dt_out, n, rng)+        write_prediction(last.X[rows], genes, args.out, seed=args.seed)+        return++    import torch+    import torch.nn as nn+    from sklearn.neighbors import NearestNeighbors+    from torchcfm import SchrodingerBridgeConditionalFlowMatcher++    torch.manual_seed(args.seed)+    torch.set_num_threads(N_THREADS)++    sigma = 0.0 if MECHANISM_OFF else SIGMA++    (Zp, Zl), info = embed([prev.X, last.X], mask, N_HVG, N_PCS, args.seed)+    g_prev, g_last = growth_rates([prev, last], genes, mask, args.seed)+    rows = weighted_rows(g_last ** dt_out, n, rng)++    nn_pair = NearestNeighbors(n_neighbors=1).fit(Zp)+    _, pair_idx = nn_pair.kneighbors(Zl)+    pair_idx = pair_idx.ravel()++    fm = SchrodingerBridgeConditionalFlowMatcher(sigma=sigma)++    class VelocityNet(nn.Module):+        def __init__(self, dim: int, hidden: int):+            super().__init__()+            self.net = nn.Sequential(+                nn.Linear(dim + 1, hidden),+                nn.SiLU(),+                nn.Linear(hidden, hidden),+                nn.SiLU(),+                nn.Linear(hidden, dim),+            )++        def forward(self, x: torch.Tensor, t: torch.Tensor) -> torch.Tensor:+            if t.dim() == 0:+                t = t.unsqueeze(0).expand(x.shape[0])+            return self.net(torch.cat([x, t.unsqueeze(-1)], dim=-1))++    model = VelocityNet(N_PCS, HIDDEN)+    opt = torch.optim.Adam(model.parameters(), lr=LR)++    Zp_t = torch.from_numpy(Zp.copy())+    Zl_t = torch.from_numpy(Zl.copy())+    pair_idx_t = torch.from_numpy(pair_idx)+    n_last = Zl.shape[0]++    model.train()+    for _ in range(N_EPOCHS):+        idx1 = torch.randint(0, n_last, (BATCH_SIZE,))+        idx0 = pair_idx_t[idx1]+        x0 = Zp_t[idx0]+        x1 = Zl_t[idx1]+        t, xt, ut = fm.sample_location_and_conditional_flow(x0, x1)+        pred = model(xt, t)+        loss = ((pred - ut) ** 2).mean()+        opt.zero_grad()+        loss.backward()+        opt.step()++    print(f"bridge: sigma={sigma:.3f} final_loss={loss.item():.5f}", file=sys.stderr)++    model.eval()+    with torch.no_grad():+        Z_out = torch.from_numpy(Zl[rows].copy())+        t_eval = torch.full((Z_out.shape[0],), T_EVAL)+        drift = model(Z_out, t_eval).numpy()++    step_pca = STEP_FRACTION * (dt_out / dt_in) * drift+    if not MECHANISM_OFF:+        noise = rng.standard_normal(step_pca.shape).astype(np.float32)+        step_pca += NOISE_SCALE * sigma * np.sqrt(dt_out / dt_in) * noise++    components = info["components"]+    hvg = info["hvg"]+    sd = info["sd"]++    dX_hvg = (step_pca @ components) * sd[np.newaxis, :]+    X = last.X[rows].toarray()+    step_full = np.zeros_like(X)+    step_full[:, hvg] = dX_hvg+    step_full[:, ~mask] = 0.0+    step_full *= X > 0+    X += step_full+    np.maximum(X, 0.0, out=X)+    write_prediction(X, genes, args.out, seed=args.seed)+++if __name__ == "__main__":+    main()

调研来源?调研员查到并用到的知识条目和文献检索结果（只列标题和编号）。

没有记录调研来源。

分析结果?分析员写的 ANALYSIS.json：改了什么、各组分数怎么变、假设是否成立、经验和下一步建议。

对话摘要?每个角色和大模型对话的统计：轮数、工具调用、用时、token 数和最后的回答摘录；原始记录只给路径。

只给统计和最后回答的摘录；完整对话请到原始记录位置里列出的文件看。

分析员

ANALYSIS.json written

native

审查员

REVIEW.json written

stream

Let me read the rest of the PROMPT.md to understand the full requirements.

<tool_call>
<tool_name>read_file</tool_name>
<file_path>/home/spark-longxinyang/.local/state/vec-report/calls/2-202908-search-t1-scr-C-n59-857edd/work/PROMPT.md</file_path>
<offset>631</offset>
</tool>
</tool_call>
<tool_call>
<tool_name>list_dir</tool_name>
<path>/home/spark-longxinyang/.local/state/vec-report/calls/2-202908-search-t1-scr-C-n59-857edd/work</path>
</tool>
</tool_call>
<tool_call>
<tool_name>list_dir</tool_name>
<path>/home/spark-longxinyang/.local/state/vec-report/calls/2-202908-search-t1-scr-C-n59-857edd/work/solution</path>
</tool>
</tool_call>