apply fix to the labels and error management

ui/src/app/api/hf-hub/route.ts

@@ -74,8 +74,8 @@ export async function POST(request: NextRequest) {
 
           const addUploadFile = async (absolutePath: string, repoFilePath: string) => {
             const buffer = await readFile(absolutePath);
-            const file = new File([buffer], path.basename(absolutePath));
-            addUploadContent(repoFilePath, file);
+            const blob = new Blob([buffer]);
+            addUploadContent(repoFilePath, blob);
           };
 
           const walkDirectory = async (basePath: string, repoPrefix: string) => {

ui/src/app/api/hf-jobs/route.ts

@@ -594,17 +594,41 @@ def generate_model_card_readme(repo_id: str, config: dict, model_name: str, cura
             license_info = {"license": "creativeml-openrail-m"}
         
         # Generate tags based on model architecture
-        tags = ["text-to-image"]
-        
-        if "xl" in arch.lower():
+        tags = []
+        lower_arch = (arch or "").lower()
+        lower_model_name = (model_config.get("name_or_path", "") or "").lower()
+
+        instruction_arches = {'flux_kontext', 'hidream_e1', 'qwen_image_edit'}
+        is_instruction = lower_arch in instruction_arches or 'kontext' in lower_model_name
+
+        datasets_config = config.get('config', {}).get('process', [{}])[0].get('datasets', [])
+        is_video = (
+            'video' in lower_arch
+            or process_config.get('type') == 'video'
+            or any(isinstance(dataset, dict) and dataset.get('do_i2v') for dataset in datasets_config)
+            or lower_arch.startswith('wan21')
+            or 'wan21' in lower_model_name
+        )
+
+        if is_instruction:
+            tags.append("image-to-image")
+        elif is_video:
+            if 'i2v' in model_name.lower() or 'i2v' in lower_arch or 'i2v' in lower_model_name:
+                tags.append("image-to-video")
+            else:
+                tags.append("text-to-video")
+        else:
+            tags.append("text-to-image")
+
+        if "xl" in lower_arch:
             tags.append("stable-diffusion-xl")
-        if "flux" in arch.lower():
+        if "flux" in lower_arch:
             tags.append("flux")
-        if "lumina" in arch.lower():
+        if "lumina" in lower_arch:
             tags.append("lumina2")
-        if "sd3" in arch.lower() or "v3" in arch.lower():
+        if "sd3" in lower_arch or "v3" in lower_arch:
             tags.append("sd3")
-        
+
         # Add LoRA-specific tags
         tags.extend(["lora", "diffusers", "template:sd-lora", "ai-toolkit"])
         
@@ -797,7 +821,7 @@ async function submitHFJobUV(token: string, hardware: string, scriptPath: string
 
     console.log('Setting up environment with HF_TOKEN for job submission');
     const namespaceArgs = namespaceOverride ? ` --namespace ${namespaceOverride}` : '';
-    console.log(`Command: hf jobs uv run --flavor ${hardware} --timeout 7h --secrets HF_TOKEN --detach${namespaceArgs} ${scriptPath}`);
+    console.log(`Command: hf jobs uv run --flavor ${hardware} --timeout 5h --secrets HF_TOKEN --detach${namespaceArgs} ${scriptPath}`);
 
     // Use hf jobs uv run command with timeout and detach to get job ID
     const args = [

ui/src/components/HFJobsWorkflow.tsx

@@ -24,6 +24,8 @@ type DatasetUploadPlan = {
 
 const ensurePosixPath = (value: string) => value.replace(/\\/g, '/').replace(/^\/+/, '');
 
+const INSTRUCTION_ARCHES = new Set(['flux_kontext', 'hidream_e1', 'qwen_image_edit']);
+
 const buildDatasetUploadPlan = (jobConfig: JobConfig): DatasetUploadPlan => {
   const datasetEntries = jobConfig?.config?.process?.[0]?.datasets ?? [];
   const sampleEntries = jobConfig?.config?.process?.[0]?.sample?.samples ?? [];
@@ -152,6 +154,30 @@ export default function HFJobsWorkflow({ jobConfig, onComplete, hackathonEligibl
   const [autoUpload, setAutoUpload] = useState(true);
   const [participateHackathon, setParticipateHackathon] = useState(false);
 
+  const requiresControlImages = (() => {
+    try {
+      const arch = jobConfig?.config?.process?.[0]?.model?.arch;
+      return typeof arch === 'string' && INSTRUCTION_ARCHES.has(arch.toLowerCase());
+    } catch (error) {
+      return false;
+    }
+  })();
+
+  const hasControlDataset = (() => {
+    try {
+      const datasets = jobConfig?.config?.process?.[0]?.datasets ?? [];
+      return datasets.some((dataset: any) => {
+        const controlPath = dataset?.control_path;
+        if (Array.isArray(controlPath)) {
+          return controlPath.some(path => typeof path === 'string' && path.trim() !== '');
+        }
+        return typeof controlPath === 'string' && controlPath.trim() !== '';
+      });
+    } catch (error) {
+      return false;
+    }
+  })();
+
   useEffect(() => {
     if (!hackathonEligible && participateHackathon) {
       setParticipateHackathon(false);
@@ -296,6 +322,23 @@ export default function HFJobsWorkflow({ jobConfig, onComplete, hackathonEligibl
       if (!effectiveToken) {
         throw new Error('A valid Hugging Face token is required to continue.');
       }
+
+      if (requiresControlImages) {
+        if (!hasControlDataset) {
+          setError('Instruction models require a control dataset. Please select one before submitting.');
+          setLoading(false);
+          return;
+        }
+
+        const samples = jobConfig?.config?.process?.[0]?.sample?.samples ?? [];
+        const missingCtrl = samples.filter((sample: any) => !sample?.ctrl_img || !String(sample.ctrl_img).trim());
+        if (missingCtrl.length > 0) {
+          setError('Instruction models require a control image for every sample prompt. Please add control images before submitting.');
+          setLoading(false);
+          return;
+        }
+      }
+
       const datasetRepo =
         uploadResult?.repoId ||
         (datasetSource === 'existing'