1919810

src/mcm26b/training_ui/schema.py

@@ -144,7 +144,8 @@ CONFIG_FIELDS: list[dict[str, Any]] = [
     {
         "key": "init", "kind": "text", "group": "paths", "optional": True,
         "label": "初始权重 --init", "default": None,
-        "help": "留空=从零；填上一阶段 .zip 即课程迁移（步数计数重置）。",
+        "help": "留空=从零；填上一阶段 .zip 即课程迁移。PPO 超参与步数计数沿用"
+                "该模型（timesteps 按增量理解，进度条按增量显示）。",
     },
     {
         "key": "local_model", "kind": "text", "group": "paths", "optional": True,

src/mcm26b/training_ui/server.py

@@ -739,7 +739,7 @@ _STATUS_VIEW = {
     "created": ("已创建", "gray", False, "命令已构造，等待 spawn"),
     "running": ("运行中", "ok", False, "子进程持有控制台输出与独立进程组"),
     "stopping": ("停止中", "warn", False, "已发送 SIGTERM，等待子进程退出"),
-    "stopped": ("已手动停止", "warn", True, "收到停止信号后退出，模型可能未保存"),
+    "stopped": ("已手动停止", "warn", True, "收到停止信号后退出；脚本会在 step 边界保存模型与元数据"),
     "exited": ("正常结束", "ok", True, "退出码 0：训练与评估流程走完"),
     "failed": ("异常结束", "bad", True, "非零退出码或启动失败，见日志尾部"),
 }
@@ -812,15 +812,24 @@ def describe_process(snapshot: dict[str, Any]) -> dict[str, Any]:
             done_steps = value
     if done_steps == 0.0:
         done_steps = _finite_float(convergence.get("latest_timesteps")) or 0.0
-    percent = (100.0 * done_steps / total_steps) if total_steps > 0 else None
+    # init/resume（课程迁移）时 num_timesteps 从上一阶段继续，进度必须按本次
+    # 增量算；start_timesteps 由训练脚本的进度 JSON 提供，缺省按 0 处理。
+    start_steps = 0.0
+    for point in metrics:
+        value = _finite_float(point.get("start_timesteps"))
+        if value is not None:
+            start_steps = value
+            break
+    trained_steps = max(0.0, done_steps - start_steps)
+    percent = (100.0 * trained_steps / total_steps) if total_steps > 0 else None
     if percent is not None:
         percent = max(0.0, min(100.0, percent))
     elapsed = float(snapshot.get("elapsed_s") or 0.0)
     rate = _latest_rate(metrics)
     remaining: float | None = None
     eta_ts: float | None = None
-    if rate and rate > 0 and total_steps > 0 and done_steps < total_steps:
-        remaining = (total_steps - done_steps) / rate
+    if rate and rate > 0 and total_steps > 0 and trained_steps < total_steps:
+        remaining = (total_steps - trained_steps) / rate
         eta_ts = float(snapshot.get("finished_at") or time.time()) + remaining
     if not terminal and remaining is None:
         cap = _finite_float(config.get("max_wall_s")) or 0.0
@@ -835,6 +844,8 @@ def describe_process(snapshot: dict[str, Any]) -> dict[str, Any]:
         "timeline": timeline,
         "percent": percent,
         "done_steps": done_steps,
+        "start_steps": start_steps,
+        "trained_steps": trained_steps,
         "total_steps": total_steps,
         "elapsed_s": round(elapsed, 3),
         "rate_per_s": rate,

src/mcm26b/training_ui/static/app.js

@@ -347,7 +347,7 @@ function renderProcess(job) {
   chip("chip-progress", "进度",
        percent === null || percent === undefined ? "—" : `${percent.toFixed(1)}%`,
        percent !== null && percent >= 100 ? "ok" : "cyan");
-  setText("steps-done", fmtInt(process.done_steps));
+  setText("steps-done", fmtInt(process.trained_steps ?? process.done_steps));
   setText("steps-total", fmtInt(process.total_steps));
   setText("progress-percent", percent === null || percent === undefined ? "—" : `${percent.toFixed(2)}%`);
   const fill = $("bar-fill");
@@ -428,7 +428,7 @@ function renderBanner(job) {
   if (job.status === "failed") {
     showBanner("bad", `训练异常结束（退出码 <b>${job.returncode === null ? "?" : job.returncode}</b>）：${escapeHtml(job.error || "见日志尾部")}。`);
   } else if (job.status === "stopped") {
-    showBanner("warn", "已按请求停止训练（SIGTERM 独立进程组），本轮模型可能未保存。");
+    showBanner("warn", "已按请求停止训练（SIGTERM 独立进程组），脚本会在 step 边界保存当前模型与元数据。");
   } else if (job.status === "exited") {
     showBanner("ok", "训练与评估流程正常结束，模型与元数据已写入输出目录。");
   } else if (process.remaining_s !== null && process.remaining_s !== undefined
@@ -708,7 +708,7 @@ $("config-form").addEventListener("submit", async (event) => {
 });
 
 $("btn-stop").addEventListener("click", async () => {
-  if (!confirm("确认停止当前训练？\n将向独立进程组发送 SIGTERM，本轮模型可能未保存。")) return;
+  if (!confirm("确认停止当前训练？\n将向独立进程组发送 SIGTERM，脚本会先保存当前模型与元数据再退出。")) return;
   const res = await fetch("/api/train/stop", {
     method: "POST", headers: { "Content-Type": "application/json" }, body: "{}",
   });