feat: add eval_enabled flag and format-first-then-split for eval dataset

studio/backend/core/training/training.py

@@ -30,6 +30,7 @@ class TrainingBackend:
         self.step_history = []
         self.eval_loss_history = []
         self.eval_step_history = []
+        self.eval_enabled = False
         self.current_theme = "light"
 
         self.trainer.add_progress_callback(self._on_progress_update)
@@ -144,6 +145,7 @@ class TrainingBackend:
             self.step_history = []
             self.eval_loss_history = []
             self.eval_step_history = []
+            self.eval_enabled = False
             import time
             output_dir = f"./outputs/{model_name.replace('/', '_')}_{int(time.time())}"
 
@@ -215,6 +217,13 @@ class TrainingBackend:
                 dataset = dataset_result
                 eval_dataset = None
 
+            # If user set eval_steps to 0, disable evaluation entirely
+            if eval_steps is not None and float(eval_steps) <= 0:
+                eval_dataset = None
+
+            # Track whether eval is enabled for status reporting
+            self.eval_enabled = eval_dataset is not None
+
             if dataset is None or self.trainer.should_stop:
                 logger.error("Failed to load dataset or stopped by user")
                 return False

studio/backend/models/training.py

@@ -96,6 +96,7 @@ class TrainingStatus(BaseModel):
         "stopped"
     ] = Field(..., description="Current phase of training pipeline")
     is_training_running: bool = Field(..., description="True if training loop is actively running")
+    eval_enabled: bool = Field(False, description="True if evaluation dataset is configured for this training run")
     message: str = Field(..., description="Human-readable status message")
     error: Optional[str] = Field(None, description="Error details if phase is 'error'")
     details: Optional[dict] = Field(None, description="Phase-specific info, e.g. {'model_size': '8B'}")

studio/backend/routes/training.py

@@ -415,6 +415,7 @@ async def get_training_status(
             job_id=job_id,
             phase=phase,
             is_training_running=is_active,
+            eval_enabled=backend.eval_enabled,
             message=status_message,
             error=error_message,
             details=details,

studio/frontend/src/features/studio/sections/charts-content.tsx

@@ -148,7 +148,9 @@ function ema(data: LossHistoryItem[], alpha: number): SmoothedLossItem[] {
 
 export function ChartsContent({
   metrics,
-}: { metrics: TrainingChartSeries }): ReactElement {
+  isTraining,
+  evalEnabled,
+}: { metrics: TrainingChartSeries; isTraining: boolean; evalEnabled: boolean }): ReactElement {
   const [smoothing, setSmoothing] = useState(0.75);
   const [showRaw, setShowRaw] = useState(true);
   const [showSmoothed, setShowSmoothed] = useState(true);
@@ -674,10 +676,10 @@ export function ChartsContent({
                   className="size-5 text-muted-foreground/50"
                 />
                 <p className="text-sm font-medium text-muted-foreground">
-                  Evaluation not configured
+                  {isTraining && evalEnabled ? "Waiting for first evaluation step…" : "Evaluation not configured"}
                 </p>
                 <p className="text-xs text-muted-foreground/60">
-                  Set eval dataset & eval_steps to track eval loss
+                  {isTraining && evalEnabled ? "Chart will appear once eval_steps is reached" : "Set eval dataset & eval_steps to track eval loss"}
                 </p>
               </div>
             </div>

studio/frontend/src/features/studio/sections/charts-section.tsx

@@ -16,6 +16,8 @@ const SKELETON_KEYS = [
 export function ChartsSection(): ReactElement | null {
   const currentStep = useTrainingRuntimeStore((state) => state.currentStep);
   const totalSteps = useTrainingRuntimeStore((state) => state.totalSteps);
+  const isTraining = useTrainingRuntimeStore((state) => state.isTrainingRunning);
+  const evalEnabled = useTrainingRuntimeStore((state) => state.evalEnabled);
   const lossHistoryRaw = useTrainingRuntimeStore((state) => state.lossHistory);
   const lrHistoryRaw = useTrainingRuntimeStore((state) => state.lrHistory);
   const gradNormHistoryRaw = useTrainingRuntimeStore(
@@ -70,7 +72,7 @@ export function ChartsSection(): ReactElement | null {
         </div>
       }
     >
-      <ChartsContent metrics={series} />
+      <ChartsContent metrics={series} isTraining={isTraining} evalEnabled={evalEnabled} />
     </Suspense>
   );
 }

studio/frontend/src/features/training/stores/training-runtime-store.ts

@@ -12,6 +12,7 @@ const initialState: TrainingRuntimeState = {
   jobId: null,
   phase: "idle",
   isTrainingRunning: false,
+  evalEnabled: false,
   message: "Ready to train",
   error: null,
   isHydrating: false,
@@ -144,6 +145,7 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
         jobId: payload.job_id || state.jobId,
         phase: payload.phase,
         isTrainingRunning: payload.is_training_running,
+        evalEnabled: payload.eval_enabled ?? state.evalEnabled,
         message: payload.message,
         error: payload.error,
         startError: null,

studio/frontend/src/features/training/types/runtime.ts

@@ -12,6 +12,7 @@ export interface TrainingStatusResponse {
   job_id: string;
   phase: TrainingPhase;
   is_training_running: boolean;
+  eval_enabled: boolean;
   message: string;
   error: string | null;
   details?: {
@@ -63,6 +64,7 @@ export interface TrainingRuntimeState {
   jobId: string | null;
   phase: TrainingPhase;
   isTrainingRunning: boolean;
+  evalEnabled: boolean;
   message: string;
   error: string | null;
   isHydrating: boolean;