修改

backend/app/api/data_management.py

 from fastapi import APIRouter, File, Form, HTTPException, Query, Response, UploadFile
-from pydantic import BaseModel, Field
 
+from app.schemas.data_management import CategoryCreateRequest, CategoryUpdateRequest, FileUpdateRequest
 from app.services.data_management_service import DataManagementService
 from app.utils.response import success_response
 
@@ -9,15 +9,6 @@ router = APIRouter()
 service = DataManagementService()
 
 
-class CategoryCreateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=100)
-    parent_id: str | int | None = Field(default='all')
-
-
-class CategoryUpdateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=100)
-
-
 @router.get('/categories')
 def get_categories():
     return success_response(data=service.get_category_tree())
@@ -91,12 +82,6 @@ def download_template():
     )
 
 
-class FileUpdateRequest(BaseModel):
-    filename: str | None = Field(default=None, min_length=1, max_length=255)
-    remark: str | None = Field(default=None, max_length=500)
-    category_id: str | None = Field(default=None)
-
-
 @router.put('/files/{file_id}')
 def update_file(file_id: str, request: FileUpdateRequest):
     try:

backend/app/api/eval_management.py

 from fastapi import APIRouter, HTTPException, Query
-from pydantic import BaseModel
 
+from app.schemas.eval_management import EvalRequest
 from app.services.eval_service import EvalService
 from app.utils.response import success_response
 
@@ -8,11 +8,6 @@ router = APIRouter()
 service = EvalService()
 
 
-class EvalRequest(BaseModel):
-    model_id: int
-    package_id: int
-
-
 @router.get('/packages')
 def list_packages(
     category_id: str = Query(default=''),

backend/app/api/monitor.py

 from fastapi import APIRouter, HTTPException, Query
-from pydantic import BaseModel, Field
 
+from app.schemas.monitor import CreateExperimentRequest
 from app.services.monitor_service import MonitorService
 from app.utils.response import success_response
 
@@ -8,33 +8,6 @@ router = APIRouter()
 service = MonitorService()
 
 
-# ── 请求 Schema ───────────────────────────────────────────────────────────────
-
-class MPCParamsSchema(BaseModel):
-    P: int = Field(default=20, ge=5, le=100, description='预测时域')
-    M: int = Field(default=5, ge=1, le=20, description='控制时域')
-    Q: float = Field(default=10.0, gt=0, description='跟踪误差权重')
-    R: float = Field(default=0.1, gt=0, description='控制增量权重')
-    alpha: float = Field(default=0.8, gt=0, lt=1, description='参考轨迹柔化系数')
-    u_min: float = Field(default=0.0, description='电流下限(A)')
-    u_max: float = Field(default=10.0, description='电流上限(A)')
-    du_min: float = Field(default=-2.0, description='电流增量下限(A/步)')
-    du_max: float = Field(default=2.0, description='电流增量上限(A/步)')
-    y_min: float = Field(default=-50.0, description='温度约束下限(°C)')
-    y_max: float = Field(default=200.0, description='温度约束上限(°C)')
-    correction_gain: float = Field(default=0.5, ge=0, le=1, description='反馈校正增益')
-
-
-class CreateExperimentRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=255)
-    model_id: int
-    input_csv_path: str = Field(min_length=1, description='读取传感器数据的CSV文件路径')
-    output_csv_path: str = Field(min_length=1, description='输出曲线数据的CSV文件路径')
-    target_temp: float = Field(description='目标温度(°C)')
-    sampling_interval: float = Field(default=1.0, gt=0, le=3600, description='采样周期(秒)')
-    mpc_params: MPCParamsSchema = Field(default_factory=MPCParamsSchema)
-
-
 # ── 元数据接口 ────────────────────────────────────────────────────────────────
 
 @router.get('/models')

backend/app/api/package_management.py

 from fastapi import APIRouter, HTTPException, Query
-from pydantic import BaseModel, Field
 
+from app.schemas.package_management import (
+    CategoryCreateRequest,
+    CategoryUpdateRequest,
+    PackageCreateRequest,
+    PackageUpdateRequest,
+    PreviewRequest,
+)
 from app.services.package_management_service import PackageManagementService
 from app.utils.response import success_response
 
@@ -9,31 +15,6 @@ router = APIRouter()
 service = PackageManagementService()
 
 
-class CategoryCreateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=100)
-    parent_id: str | int | None = Field(default='all')
-
-
-class CategoryUpdateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=100)
-
-
-class CleanRules(BaseModel):
-    enabled: bool = False
-    newton_interp: bool = False
-    current_min: float | None = None
-    current_max: float | None = None
-    voltage_min: float | None = None
-    voltage_max: float | None = None
-    temperature_min: float | None = None
-    temperature_max: float | None = None
-
-
-class SmoothConfig(BaseModel):
-    enabled: bool = False
-    window: int = Field(default=5, ge=2, le=500)
-
-
 @router.get('/categories')
 def get_categories():
     return success_response(data=service.get_category_tree())
@@ -81,32 +62,6 @@ def list_all_data_files(
     ))
 
 
-class PackageUpdateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=255)
-    category_id: str | int | None = Field(default=None)
-    remark: str | None = Field(default=None)
-
-
-class PackageCreateRequest(BaseModel):
-    name: str = Field(min_length=1, max_length=255)
-    category_id: str | int | None = Field(default=None)
-    remark: str | None = Field(default=None)
-    file_ids: list[int] = Field(default_factory=list)
-    clean_rules: CleanRules | None = Field(default=None)
-    smooth: SmoothConfig | None = Field(default=None)
-    auto_split: bool = Field(default=False)
-    row_start: int | None = Field(default=None, ge=1)
-    row_end: int | None = Field(default=None, ge=1)
-
-
-class PreviewRequest(BaseModel):
-    file_ids: list[int] = Field(default_factory=list)
-    clean_rules: CleanRules | None = Field(default=None)
-    smooth: SmoothConfig | None = Field(default=None)
-    row_start: int | None = Field(default=None, ge=1)
-    row_end: int | None = Field(default=None, ge=1)
-
-
 @router.post('/preview')
 def preview_package(request: PreviewRequest):
     try:

backend/app/api/train_management.py

 from fastapi import APIRouter, HTTPException, Query
-from pydantic import BaseModel, Field
 
+from app.schemas.train_management import CreateTaskRequest, SaveModelRequest, UpdateModelRequest
 from app.services.train_service import TrainService
 from app.utils.response import success_response
 
@@ -8,34 +8,6 @@ router = APIRouter()
 service = TrainService()
 
 
-# ── request schemas ───────────────────────────────────────────────────────────
-
-class LSTMParams(BaseModel):
-    seq_len: int = Field(default=20, ge=5, le=500)
-    hidden_size: int = Field(default=64, ge=8, le=1024)
-    num_layers: int = Field(default=2, ge=1, le=8)
-    epochs: int = Field(default=50, ge=1, le=2000)
-    batch_size: int = Field(default=32, ge=1, le=512)
-    learning_rate: float = Field(default=0.001, gt=0, le=1)
-
-
-class CreateTaskRequest(BaseModel):
-    model_name: str = Field(min_length=1, max_length=255)
-    train_package_id: int
-    val_package_id: int
-    params: LSTMParams = Field(default_factory=LSTMParams)
-
-
-class SaveModelRequest(BaseModel):
-    model_name: str | None = Field(default=None, max_length=255)
-    description: str | None = None
-
-
-class UpdateModelRequest(BaseModel):
-    model_name: str | None = Field(default=None, max_length=255)
-    description: str | None = None
-
-
 # ── packages ──────────────────────────────────────────────────────────────────
 
 @router.get('/packages')

backend/app/ml/lstm_trainer.py

@@ -97,7 +97,8 @@ def train_lstm(
         train_records: list of dicts for the training set.
         val_records:   list of dicts for the validation set.
         params:        hyper-parameter dict (seq_len, hidden_size, num_layers,
-                       epochs, batch_size, learning_rate).
+                       epochs, batch_size, learning_rate,
+                       lr_factor, lr_patience, lr_min).
         save_path:     destination .pt file.
         on_progress:   callback(pct, epoch, train_loss, val_loss) called after each epoch.
         cancel_event:  when set, training stops with InterruptedError.
@@ -114,15 +115,20 @@ def train_lstm(
     batch_size = max(1, int(params.get('batch_size', 32)))
     lr = float(params.get('learning_rate', 0.001))
 
+    # ── scheduler 超参数 ──────────────────────────────────────────────────────
+    lr_factor = float(params.get('lr_factor', 0.5))
+    lr_patience = max(1, int(params.get('lr_patience', 5)))
+    lr_min = float(params.get('lr_min', 1e-6))
+
     # ── data preparation ────────────────────────────────────────────────────
     train_data = _extract_features(train_records)
-    val_data   = _extract_features(val_records)
+    val_data = _extract_features(val_records)
 
     min_required = seq_len + 10
     if len(train_data) < min_required:
         raise ValueError(
-            f'\u8bad\u7ec3\u96c6\u6709\u6548\u6570\u636e\u91cf\u4e0d\u8db3\uff1a\u9700\u81f3\u5c11 {min_required} \u6761\uff0c\u5f53\u524d\u4ec5 {len(train_data)} \u6761\u3002'
-            '\u8bf7\u68c0\u67e5\u6570\u636e\u5305\u5185\u5bb9\u6216\u51cf\u5c0f\u5e8f\u5217\u957f\u5ea6\u3002'
+            f'训练集有效数据量不足：需至少 {min_required} 条，当前仅 {len(train_data)} 条。'
+            '请检查数据包内容或减小序列长度。'
         )
 
     # min-max normalisation fitted on train set only
@@ -161,134 +167,15 @@ def train_lstm(
     optimizer = torch.optim.Adam(model.parameters(), lr=lr)
     criterion = nn.MSELoss()
 
-    train_loss = 0.0
-    val_loss: float | None = None
-
-    for epoch in range(epochs):
-        if cancel_event.is_set():
-            raise InterruptedError('\u8bad\u7ec3\u5df2\u53d6\u6d88')
-
-        # ── train step ───────────────────────────────────────────────────────
-        model.train()
-        epoch_loss = 0.0
-        for xb, yb in train_loader:
-            optimizer.zero_grad()
-            pred = model(xb)
-            loss = criterion(pred, yb)
-            loss.backward()
-            optimizer.step()
-            epoch_loss += loss.item() * len(xb)
-
-        train_loss = epoch_loss / len(X_train)
-
-        # ── val step ─────────────────────────────────────────────────────────
-        if has_val:
-            model.eval()
-            with torch.no_grad():
-                val_pred = model(X_val_t)
-                val_loss = criterion(val_pred, y_val_t).item()
-
-        pct = int((epoch + 1) / epochs * 100)
-        on_progress(pct, epoch + 1, train_loss, val_loss)
-
-    # ── persist ──────────────────────────────────────────────────────────────
-    save_path.parent.mkdir(parents=True, exist_ok=True)
-    torch.save(
-        {
-            'model_state': model.state_dict(),
-            'params': params,
-            'data_min': data_min.tolist(),
-            'data_max': data_max.tolist(),
-            'feature_cols': FEATURE_COLS,
-            'target_col': TARGET_COL,
-            'input_size': input_size,
-            'hidden_size': hidden_size,
-            'num_layers': num_layers,
-            'seq_len': seq_len,
-        },
-        save_path,
-    )
-
-    return {
-        'train_loss': round(float(train_loss), 6),
-        'val_loss':   round(float(val_loss),   6) if val_loss is not None else None,
-    }
-
-    """
-    Train an LSTM model on *records* and persist it to *save_path*.
-
-    Args:
-        records:      list of dicts with keys in FEATURE_COLS.
-        params:       hyper-parameter dict (seq_len, hidden_size, num_layers,
-                      epochs, batch_size, learning_rate).
-                      train_ratio is ignored – fixed 70/15/15 split is used.
-        save_path:    destination .pt file.
-        on_progress:  callback(pct, train_loss, val_loss) called after each epoch.
-        cancel_event: when set, training stops with InterruptedError.
-
-    Returns:
-        {'train_loss': float, 'val_loss': float|None, 'test_loss': float|None}
-    """
-    _check_torch()
-
-    seq_len = max(1, int(params.get('seq_len', 20)))
-    hidden_size = max(1, int(params.get('hidden_size', 64)))
-    num_layers = max(1, int(params.get('num_layers', 2)))
-    epochs = max(1, int(params.get('epochs', 50)))
-    batch_size = max(1, int(params.get('batch_size', 32)))
-    lr = float(params.get('learning_rate', 0.001))
-
-    # ── data preparation ────────────────────────────────────────────────────
-    data = _extract_features(records)
-    min_required = seq_len + 10
-    if len(data) < min_required:
-        raise ValueError(
-            f'有效数据量不足：需至少 {min_required} 条，当前仅 {len(data)} 条。'
-            '请检查数据包内容或减小序列长度。'
-        )
-
-    # min-max normalisation per feature
-    data_min = data.min(axis=0)
-    data_max = data.max(axis=0)
-    data_range = data_max - data_min
-    data_range[data_range == 0] = 1.0
-    data_norm = (data - data_min) / data_range
-
-    X, y = _make_sequences(data_norm, seq_len)
-
-    # Fixed 70 / 15 / 15 split
-    n_total = len(X)
-    n_train = max(1, int(n_total * _TRAIN_RATIO))
-    n_val   = max(1, int(n_total * _VAL_RATIO))
-    # test gets the remainder so the three parts always sum to n_total
-    X_train, y_train = X[:n_train], y[:n_train]
-    X_val,   y_val   = X[n_train : n_train + n_val], y[n_train : n_train + n_val]
-    X_test,  y_test  = X[n_train + n_val :],          y[n_train + n_val :]
-
-    device = torch.device('cpu')
-    X_train_t = torch.tensor(X_train).to(device)
-    y_train_t = torch.tensor(y_train).to(device)
-    has_val  = len(X_val) > 0
-    has_test = len(X_test) > 0
-    if has_val:
-        X_val_t = torch.tensor(X_val).to(device)
-        y_val_t = torch.tensor(y_val).to(device)
-    if has_test:
-        X_test_t = torch.tensor(X_test).to(device)
-        y_test_t = torch.tensor(y_test).to(device)
-
-    train_loader = DataLoader(
-        TensorDataset(X_train_t, y_train_t),
-        batch_size=batch_size,
-        shuffle=True,
+    # ── scheduler ────────────────────────────────────────────────────────────
+    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
+        optimizer,
+        mode='min',
+        factor=lr_factor,
+        patience=lr_patience,
+        min_lr=lr_min,
     )
 
-    # ── model ────────────────────────────────────────────────────────────────
-    input_size = len(FEATURE_COLS)
-    model = _LSTMModel(input_size, hidden_size, num_layers).to(device)
-    optimizer = torch.optim.Adam(model.parameters(), lr=lr)
-    criterion = nn.MSELoss()
-
     train_loss = 0.0
     val_loss: float | None = None
 
@@ -316,16 +203,11 @@ def train_lstm(
                 val_pred = model(X_val_t)
                 val_loss = criterion(val_pred, y_val_t).item()
 
-        pct = int((epoch + 1) / epochs * 100)
-        on_progress(pct, train_loss, val_loss)
+            # 根据 val_loss 动态调整学习率
+            scheduler.step(val_loss)
 
-    # ── test-set evaluation (completely hidden during training) ──────────────
-    test_loss: float | None = None
-    if has_test:
-        model.eval()
-        with torch.no_grad():
-            test_pred = model(X_test_t)
-            test_loss = criterion(test_pred, y_test_t).item()
+        pct = int((epoch + 1) / epochs * 100)
+        on_progress(pct, epoch + 1, train_loss, val_loss)
 
     # ── persist ──────────────────────────────────────────────────────────────
     save_path.parent.mkdir(parents=True, exist_ok=True)
@@ -347,6 +229,6 @@ def train_lstm(
 
     return {
         'train_loss': round(float(train_loss), 6),
-        'val_loss':   round(float(val_loss),   6) if val_loss   is not None else None,
-        'test_loss':  round(float(test_loss),  6) if test_loss  is not None else None,
+        'val_loss':   round(float(val_loss),   6) if val_loss is not None else None,
     }
+

backend/app/schemas/data_management.py

+from pydantic import BaseModel, Field
+
+
+class CategoryCreateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=100)
+    parent_id: str | int | None = Field(default='all')
+
+
+class CategoryUpdateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=100)
+
+
+class FileUpdateRequest(BaseModel):
+    filename: str | None = Field(default=None, min_length=1, max_length=255)
+    remark: str | None = Field(default=None, max_length=500)
+    category_id: str | None = Field(default=None)

backend/app/schemas/eval_management.py

+from pydantic import BaseModel
+
+
+class EvalRequest(BaseModel):
+    model_id: int
+    package_id: int

backend/app/schemas/monitor.py

+from pydantic import BaseModel, Field
+
+
+class MPCParamsSchema(BaseModel):
+    P: int = Field(default=20, ge=5, le=100, description='预测时域')
+    M: int = Field(default=5, ge=1, le=20, description='控制时域')
+    Q: float = Field(default=10.0, gt=0, description='跟踪误差权重')
+    R: float = Field(default=0.1, gt=0, description='控制增量权重')
+    alpha: float = Field(default=0.8, gt=0, lt=1, description='参考轨迹柔化系数')
+    u_min: float = Field(default=0.0, description='电流下限(A)')
+    u_max: float = Field(default=10.0, description='电流上限(A)')
+    du_min: float = Field(default=-2.0, description='电流增量下限(A/步)')
+    du_max: float = Field(default=2.0, description='电流增量上限(A/步)')
+    y_min: float = Field(default=-50.0, description='温度约束下限(°C)')
+    y_max: float = Field(default=200.0, description='温度约束上限(°C)')
+    correction_gain: float = Field(default=0.5, ge=0, le=1, description='反馈校正增益')
+
+
+class CreateExperimentRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=255)
+    model_id: int
+    input_csv_path: str = Field(min_length=1, description='读取传感器数据的CSV文件路径')
+    output_csv_path: str = Field(min_length=1, description='输出曲线数据的CSV文件路径')
+    target_temp: float = Field(description='目标温度(°C)')
+    sampling_interval: float = Field(default=1.0, gt=0, le=3600, description='采样周期(秒)')
+    mpc_params: MPCParamsSchema = Field(default_factory=MPCParamsSchema)

backend/app/schemas/package_management.py

+from pydantic import BaseModel, Field
+
+
+class CategoryCreateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=100)
+    parent_id: str | int | None = Field(default='all')
+
+
+class CategoryUpdateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=100)
+
+
+class CleanRules(BaseModel):
+    enabled: bool = False
+    newton_interp: bool = False
+    current_min: float | None = None
+    current_max: float | None = None
+    voltage_min: float | None = None
+    voltage_max: float | None = None
+    temperature_min: float | None = None
+    temperature_max: float | None = None
+
+
+class SmoothConfig(BaseModel):
+    enabled: bool = False
+    window: int = Field(default=5, ge=2, le=500)
+
+
+class PackageUpdateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=255)
+    category_id: str | int | None = Field(default=None)
+    remark: str | None = Field(default=None)
+
+
+class PackageCreateRequest(BaseModel):
+    name: str = Field(min_length=1, max_length=255)
+    category_id: str | int | None = Field(default=None)
+    remark: str | None = Field(default=None)
+    file_ids: list[int] = Field(default_factory=list)
+    clean_rules: CleanRules | None = Field(default=None)
+    smooth: SmoothConfig | None = Field(default=None)
+    auto_split: bool = Field(default=False)
+    row_start: int | None = Field(default=None, ge=1)
+    row_end: int | None = Field(default=None, ge=1)
+
+
+class PreviewRequest(BaseModel):
+    file_ids: list[int] = Field(default_factory=list)
+    clean_rules: CleanRules | None = Field(default=None)
+    smooth: SmoothConfig | None = Field(default=None)
+    row_start: int | None = Field(default=None, ge=1)
+    row_end: int | None = Field(default=None, ge=1)

backend/app/schemas/train_management.py

+from pydantic import BaseModel, Field
+
+
+class LSTMParams(BaseModel):
+    seq_len: int = Field(default=20, ge=5, le=500)
+    hidden_size: int = Field(default=64, ge=8, le=1024)
+    num_layers: int = Field(default=2, ge=1, le=8)
+    epochs: int = Field(default=50, ge=1, le=2000)
+    batch_size: int = Field(default=32, ge=1, le=512)
+    learning_rate: float = Field(default=0.001, gt=0, le=1)
+
+
+class CreateTaskRequest(BaseModel):
+    model_name: str = Field(min_length=1, max_length=255)
+    train_package_id: int
+    val_package_id: int
+    params: LSTMParams = Field(default_factory=LSTMParams)
+
+
+class SaveModelRequest(BaseModel):
+    model_name: str | None = Field(default=None, max_length=255)
+    description: str | None = None
+
+
+class UpdateModelRequest(BaseModel):
+    model_name: str | None = Field(default=None, max_length=255)
+    description: str | None = None

backend/uploads/test/test.csv

@@ -27,3 +27,15 @@
 25,20.534,0.1796,23.4272,35.0
 26,20.534,0.1796,23.4272,35.0
 27,20.534,0.1949,23.4272,35.0
+28,20.534,0.0,23.4272,35.0
+29,20.534,0.1796,23.4272,35.0
+30,20.534,0.1796,23.4272,35.0
+31,20.534,0.0,23.4272,35.0
+32,20.534,0.1796,23.4272,35.0
+33,20.534,0.1796,23.4272,35.0
+34,20.534,0.1949,23.4272,35.0
+35,20.534,0.1949,23.4272,35.0
+36,20.534,0.1949,23.4272,35.0
+37,20.534,0.1949,23.4272,35.0
+38,20.534,0.1949,23.4272,35.0
+39,20.534,0.1354,23.4272,35.0

frontend/src/components/common/UnderConstruction.vue

-<script setup>
-defineProps({
-  title: {
-    type: String,
-    default: '页面建设中',
-  },
-})
-</script>
-
-<template>
-  <section class="uc-page">
-    <div class="uc-card">
-      <h2>{{ title }}</h2>
-      <p>该模块正在建设中，可先使用“数据管理”完成数据上传与查看。</p>
-    </div>
-  </section>
-</template>
-
-<style lang="scss" scoped>
-.uc-page {
-  height: 100%;
-  display: grid;
-  place-items: center;
-  background: var(--bg-page);
-}
-
-.uc-card {
-  width: min(480px, 100%);
-  border-radius: 4px;
-  padding: 40px 32px;
-  background: var(--bg-white);
-  border: 1px solid var(--border-color);
-  box-shadow: var(--shadow-card);
-  text-align: center;
-
-  h2 {
-    margin: 0 0 12px;
-    font-size: 16px;
-    font-weight: 600;
-    color: var(--text-primary);
-  }
-
-  p {
-    margin: 0;
-    color: var(--text-secondary);
-    font-size: 13px;
-    line-height: 1.6;
-  }
-}
-</style>