diff --git a/acestep/ui/gradio/i18n/en.json b/acestep/ui/gradio/i18n/en.json
index 3e33d12f..652d693d 100644
--- a/acestep/ui/gradio/i18n/en.json
+++ b/acestep/ui/gradio/i18n/en.json
@@ -5,8 +5,8 @@
   },
   "common": {
     "language_metadata": {
-    "name": "English",
-    "native_name": "English"
+      "name": "English",
+      "native_name": "English"
     }
   },
   "dataset": {
@@ -424,25 +424,55 @@
     "export_path": "Export Path",
     "export_lora_btn": "📦 Export LoRA",
     "export_status": "Export Status",
-    "stop_no_training": "\u2139 No training in progress",
-    "stop_stopping": "\u23F9\uFE0F Stopping training...",
+    "stop_no_training": "ℹ No training in progress",
+    "stop_stopping": "⏹️ Stopping training...",
     "latest_auto": "Latest (auto)",
-    "export_path_required": "\u274C Please enter an export path",
-    "invalid_lora_output_dir": "\u274C Invalid LoRA output directory",
-    "no_checkpoints_found": "\u274C No checkpoints found",
-    "no_trained_model_found": "\u274C No trained model found in {path}",
-    "invalid_export_path": "\u274C Invalid export path",
-    "lora_exported": "\u2705 LoRA exported to {path}",
-    "export_failed": "\u274C Export failed: {error}",
-    "lokr_output_dir_required": "\u26A0\uFE0F Enter LoKr output directory first",
-    "lokr_no_checkpoints_use_latest": "\u2139 No checkpoints found; export will use latest available weights",
-    "lokr_no_exportable_checkpoints": "\u2139 No exportable epoch checkpoints found",
-    "lokr_found_checkpoints": "\u2705 Found {count} LoKr checkpoints",
-    "lokr_selected_epoch_not_found": "\u274C Selected epoch not found: {chosen}. Available: {available}",
-    "lokr_no_weights_selected_epoch": "\u274C No LoKr weights found for selected epoch: {epoch}",
-    "lokr_no_weights_latest_checkpoint": "\u274C No LoKr weights found in latest checkpoint: {checkpoint}",
-    "lokr_no_trained_weights_found": "\u274C No trained LoKr weights found in {path}",
-    "lokr_exported": "\u2705 LoKr exported to {path}"
+    "export_path_required": "❌ Please enter an export path",
+    "invalid_lora_output_dir": "❌ Invalid LoRA output directory",
+    "no_checkpoints_found": "❌ No checkpoints found",
+    "no_trained_model_found": "❌ No trained model found in {path}",
+    "invalid_export_path": "❌ Invalid export path",
+    "lora_exported": "✅ LoRA exported to {path}",
+    "export_failed": "❌ Export failed: {error}",
+    "lokr_output_dir_required": "⚠️ Enter LoKr output directory first",
+    "lokr_no_checkpoints_use_latest": "ℹ No checkpoints found; export will use latest available weights",
+    "lokr_no_exportable_checkpoints": "ℹ No exportable epoch checkpoints found",
+    "lokr_found_checkpoints": "✅ Found {count} LoKr checkpoints",
+    "lokr_selected_epoch_not_found": "❌ Selected epoch not found: {chosen}. Available: {available}",
+    "lokr_no_weights_selected_epoch": "❌ No LoKr weights found for selected epoch: {epoch}",
+    "lokr_no_weights_latest_checkpoint": "❌ No LoKr weights found in latest checkpoint: {checkpoint}",
+    "lokr_no_trained_weights_found": "❌ No trained LoKr weights found in {path}",
+    "lokr_exported": "✅ LoKr exported to {path}",
+    "tab_train_lokr": "🚀 Train LoKr",
+    "train_section_tensors": "Preprocessed Tensors",
+    "train_section_lora": "LoRA Settings",
+    "train_section_params": "Training Parameters",
+    "lokr_section_tensors": "Preprocessed Tensors",
+    "lokr_section_settings": "LoKr Settings",
+    "lokr_tensor_selection_desc": "Select the directory containing preprocessed tensor files (`.pt` files).\nThese are created using the 'Preprocess' button in the 'Dataset Builder' tab.",
+    "lokr_linear_dim": "LoKr Linear Dim",
+    "lokr_linear_dim_info": "Rank (dimension) for LoKr adaptation matrices.",
+    "lokr_linear_alpha": "LoKr Linear Alpha",
+    "lokr_linear_alpha_info": "Scaling factor for LoKr (usually similar to dim).",
+    "lokr_factor": "LoKr Factor",
+    "lokr_factor_info": "Kronecker factor (-1 for auto).",
+    "lokr_decompose_both": "Decompose Both Sides",
+    "lokr_decompose_both_info": "When enabled, decomposes both left and right matrices.",
+    "lokr_use_tucker": "Use Tucker Decomposition",
+    "lokr_use_tucker_info": "Apply Tucker decomposition when applicable.",
+    "lokr_use_scalar": "Use Scalar Gate",
+    "lokr_use_scalar_info": "Enable scalar gating for LoKr weights.",
+    "lokr_weight_decompose": "Weight Decompose (WD)",
+    "lokr_weight_decompose_info": "Enable weight decomposition for more stable LoKr training.",
+    "lokr_learning_rate_info": "LoKr commonly uses a higher LR than LoRA. Tune per dataset.",
+    "lokr_output_dir_info": "Directory to save trained LoKr weights.",
+    "start_lokr_training_btn": "🚀 Start Training LoKr",
+    "lokr_training_loss_title": "LoKr Training Loss",
+    "lokr_export_header": "Export LoKr",
+    "export_lokr_btn": "📦 Export LoKr",
+    "lokr_checkpoint_epoch": "Checkpoint Epoch",
+    "lokr_checkpoint_epoch_info": "Select a specific epoch checkpoint to export, or keep Latest (auto).",
+    "refresh_epochs_btn": "↻ Refresh Epochs"
   },
   "help": {
     "btn_label": "?",
diff --git a/acestep/ui/gradio/i18n/he.json b/acestep/ui/gradio/i18n/he.json
index db2b8a7f..b5cf715e 100644
--- a/acestep/ui/gradio/i18n/he.json
+++ b/acestep/ui/gradio/i18n/he.json
@@ -442,7 +442,37 @@
     "lokr_no_weights_selected_epoch": "❌ לא נמצאו משקלי LoKr לאיטרציה שנבחרה: {epoch}",
     "lokr_no_weights_latest_checkpoint": "❌ לא נמצאו משקלי LoKr בנקודת הביקורת האחרונה",
     "lokr_no_trained_weights_found": "❌ לא נמצאו משקלי LoKr מאומנים ב-{path}",
-    "lokr_exported": "✅ LoKr יוצא ל-{path}"
+    "lokr_exported": "✅ LoKr יוצא ל-{path}",
+    "tab_train_lokr": "🚀 אימון LoKr",
+    "train_section_tensors": "Preprocessed Tensors",
+    "train_section_lora": "LoRA Settings",
+    "train_section_params": "Training Parameters",
+    "lokr_section_tensors": "Preprocessed Tensors",
+    "lokr_section_settings": "LoKr Settings",
+    "lokr_tensor_selection_desc": "בחר את הספרייה המכילה את קבצי הטנזורים המעובדים (`.pt`).",
+    "lokr_linear_dim": "LoKr Linear Dim",
+    "lokr_linear_dim_info": "Rank (dimension) for LoKr adaptation matrices.",
+    "lokr_linear_alpha": "LoKr Linear Alpha",
+    "lokr_linear_alpha_info": "Scaling factor for LoKr (usually similar to dim).",
+    "lokr_factor": "LoKr Factor",
+    "lokr_factor_info": "Kronecker factor (-1 for auto).",
+    "lokr_decompose_both": "Decompose Both Sides",
+    "lokr_decompose_both_info": "When enabled, decomposes both left and right matrices.",
+    "lokr_use_tucker": "Use Tucker Decomposition",
+    "lokr_use_tucker_info": "Apply Tucker decomposition when applicable.",
+    "lokr_use_scalar": "Use Scalar Gate",
+    "lokr_use_scalar_info": "Enable scalar gating for LoKr weights.",
+    "lokr_weight_decompose": "Weight Decompose (WD)",
+    "lokr_weight_decompose_info": "Enable weight decomposition for more stable LoKr training.",
+    "lokr_learning_rate_info": "LoKr commonly uses a higher LR than LoRA. Tune per dataset.",
+    "lokr_output_dir_info": "ספרייה לשמירת משקלי LoKr המאומנים.",
+    "start_lokr_training_btn": "🚀 התחל אימון LoKr",
+    "lokr_training_loss_title": "LoKr אובדן אימון (Loss)",
+    "lokr_export_header": "ייצוא LoKr",
+    "export_lokr_btn": "📦 ייצוא LoKr",
+    "lokr_checkpoint_epoch": "Checkpoint Epoch",
+    "lokr_checkpoint_epoch_info": "Select a specific epoch checkpoint to export, or keep Latest (auto).",
+    "refresh_epochs_btn": "↻ רענון Epochs"
   },
   "help": {
     "btn_label": "?",
diff --git a/acestep/ui/gradio/i18n/ja.json b/acestep/ui/gradio/i18n/ja.json
index beb6d446..34bae681 100644
--- a/acestep/ui/gradio/i18n/ja.json
+++ b/acestep/ui/gradio/i18n/ja.json
@@ -5,8 +5,8 @@
   },
   "common": {
     "language_metadata": {
-    "name": "Japanese",
-    "native_name": "日本語"
+      "name": "Japanese",
+      "native_name": "日本語"
     }
   },
   "dataset": {
@@ -426,7 +426,34 @@
     "lokr_no_weights_selected_epoch": "❌ 選択したエポック {epoch} にLoKr重みが見つかりません",
     "lokr_no_weights_latest_checkpoint": "❌ 最新チェックポイント {checkpoint} にLoKr重みが見つかりません",
     "lokr_no_trained_weights_found": "❌ {path} にトレーニング済みLoKr重みが見つかりません",
-    "lokr_exported": "✅ LoKr を {path} にエクスポートしました"
+    "lokr_exported": "✅ LoKr を {path} にエクスポートしました",
+    "tab_train_lokr": "🚀 LoKr をトレーニング",
+    "lokr_section_tensors": "前処理済みデータセット選択",
+    "lokr_section_settings": "LoKr 設定",
+    "lokr_tensor_selection_desc": "前処理されたテンソルファイル（`.pt` ファイル）を含むディレクトリを選択します。\nこれらは「データセットビルダー」タブの「前処理」ボタンを使用して作成されます。",
+    "lokr_linear_dim": "LoKr Linear Dim",
+    "lokr_linear_dim_info": "Rank (dimension) for LoKr adaptation matrices.",
+    "lokr_linear_alpha": "LoKr Linear Alpha",
+    "lokr_linear_alpha_info": "Scaling factor for LoKr (usually similar to dim).",
+    "lokr_factor": "LoKr Factor",
+    "lokr_factor_info": "Kronecker factor (-1 for auto).",
+    "lokr_decompose_both": "Decompose Both Sides",
+    "lokr_decompose_both_info": "When enabled, decomposes both left and right matrices.",
+    "lokr_use_tucker": "Use Tucker Decomposition",
+    "lokr_use_tucker_info": "Apply Tucker decomposition when applicable.",
+    "lokr_use_scalar": "Use Scalar Gate",
+    "lokr_use_scalar_info": "Enable scalar gating for LoKr weights.",
+    "lokr_weight_decompose": "Weight Decompose (WD)",
+    "lokr_weight_decompose_info": "Enable weight decomposition for more stable LoKr training.",
+    "lokr_learning_rate_info": "LoKr commonly uses a higher LR than LoRA. Tune per dataset.",
+    "lokr_output_dir_info": "トレーニング済み LoKr 重みの保存先ディレクトリ。",
+    "start_lokr_training_btn": "🚀 トレーニング開始 LoKr",
+    "lokr_training_loss_title": "LoKr トレーニング損失",
+    "lokr_export_header": "LoKr をエクスポート",
+    "export_lokr_btn": "📦 LoKr をエクスポート",
+    "lokr_checkpoint_epoch": "Checkpoint Epoch",
+    "lokr_checkpoint_epoch_info": "Select a specific epoch checkpoint to export, or keep Latest (auto).",
+    "refresh_epochs_btn": "↻ 更新 Epochs"
   },
   "help": {
     "btn_label": "?",
diff --git a/acestep/ui/gradio/i18n/zh.json b/acestep/ui/gradio/i18n/zh.json
index 3a2c5716..74f28b3e 100644
--- a/acestep/ui/gradio/i18n/zh.json
+++ b/acestep/ui/gradio/i18n/zh.json
@@ -5,8 +5,8 @@
   },
   "common": {
     "language_metadata": {
-    "name": "Chinese",
-    "native_name": "中文"
+      "name": "Chinese",
+      "native_name": "中文"
     }
   },
   "dataset": {
@@ -426,7 +426,34 @@
     "lokr_no_weights_selected_epoch": "❌ 所选轮次 {epoch} 中未找到 LoKr 权重",
     "lokr_no_weights_latest_checkpoint": "❌ 最新检查点 {checkpoint} 中未找到 LoKr 权重",
     "lokr_no_trained_weights_found": "❌ 在 {path} 中未找到训练好的 LoKr 权重",
-    "lokr_exported": "✅ LoKr 已导出到 {path}"
+    "lokr_exported": "✅ LoKr 已导出到 {path}",
+    "tab_train_lokr": "🚀 训练 LoKr",
+    "lokr_section_tensors": "预处理数据集选择",
+    "lokr_section_settings": "LoKr 设置",
+    "lokr_tensor_selection_desc": "选择包含预处理张量文件（`.pt` 文件）的目录。\n这些文件在「数据集构建」标签页中使用「预处理」按钮创建。",
+    "lokr_linear_dim": "LoKr Linear Dim",
+    "lokr_linear_dim_info": "Rank (dimension) for LoKr adaptation matrices.",
+    "lokr_linear_alpha": "LoKr Linear Alpha",
+    "lokr_linear_alpha_info": "Scaling factor for LoKr (usually similar to dim).",
+    "lokr_factor": "LoKr Factor",
+    "lokr_factor_info": "Kronecker factor (-1 for auto).",
+    "lokr_decompose_both": "Decompose Both Sides",
+    "lokr_decompose_both_info": "When enabled, decomposes both left and right matrices.",
+    "lokr_use_tucker": "Use Tucker Decomposition",
+    "lokr_use_tucker_info": "Apply Tucker decomposition when applicable.",
+    "lokr_use_scalar": "Use Scalar Gate",
+    "lokr_use_scalar_info": "Enable scalar gating for LoKr weights.",
+    "lokr_weight_decompose": "Weight Decompose (WD)",
+    "lokr_weight_decompose_info": "Enable weight decomposition for more stable LoKr training.",
+    "lokr_learning_rate_info": "LoKr commonly uses a higher LR than LoRA. Tune per dataset.",
+    "lokr_output_dir_info": "保存训练后 LoKr 权重的目录。",
+    "start_lokr_training_btn": "🚀 开始训练 LoKr",
+    "lokr_training_loss_title": "LoKr 训练损失",
+    "lokr_export_header": "导出 LoKr",
+    "export_lokr_btn": "📦 导出 LoKr",
+    "lokr_checkpoint_epoch": "Checkpoint Epoch",
+    "lokr_checkpoint_epoch_info": "Select a specific epoch checkpoint to export, or keep Latest (auto).",
+    "refresh_epochs_btn": "↻ 刷新 Epochs"
   },
   "help": {
     "btn_label": "?",
diff --git a/acestep/ui/gradio/interfaces/training.py b/acestep/ui/gradio/interfaces/training.py
index 4512bb4b..c207a77e 100644
--- a/acestep/ui/gradio/interfaces/training.py
+++ b/acestep/ui/gradio/interfaces/training.py
@@ -1,835 +1,68 @@
-"""
-Gradio UI Training Tab Module
+"""Gradio UI training-tab facade that composes focused tab builders."""
 
-Contains the dataset builder and LoRA training interface components.
-The outer gr.Tab wrapper is now created in __init__.py.
-"""
+from __future__ import annotations
 
-import os
 import gradio as gr
-from acestep.ui.gradio.i18n import t
-from acestep.ui.gradio.help_content import create_help_button
+
 from acestep.constants import DEBUG_TRAINING
+from acestep.ui.gradio.interfaces.training_dataset_builder_tab import (
+    create_dataset_builder_tab,
+)
+from acestep.ui.gradio.interfaces.training_lokr_tab import create_training_lokr_tab
+from acestep.ui.gradio.interfaces.training_lora_tab import create_training_lora_tab
+
+
+def _resolve_epoch_slider_defaults() -> tuple[int, int, int]:
+    """Return epoch slider defaults adjusted for training debug mode."""
+
+    debug_training_enabled = str(DEBUG_TRAINING).strip().upper() != "OFF"
+    if debug_training_enabled:
+        return 1, 1, 1
+    return 100, 100, 1000
 
 
 def create_training_section(dit_handler, llm_handler, init_params=None) -> dict:
-    """Create the training tab content (without the outer gr.Tab wrapper).
-    
-    The outer gr.Tab is now created in __init__.py so that Generation and
-    Training tabs are siblings under the same gr.Tabs container.
-    
+    """Create the training-tab content without the outer ``gr.Tab`` wrapper.
+
     Args:
-        dit_handler: DiT handler instance
-        llm_handler: LLM handler instance
-        init_params: Dictionary containing initialization parameters and state.
-                    If None, service will not be pre-initialized.
-        
+        dit_handler: DiT handler instance.
+        llm_handler: LLM handler instance.
+        init_params: Optional initialization parameters.
+
     Returns:
-        Dictionary of Gradio components for event handling
+        Mapping of component keys to Gradio components for training-event wiring.
     """
-    debug_training_enabled = str(DEBUG_TRAINING).strip().upper() != "OFF"
-    epoch_min = 1 if debug_training_enabled else 100
-    epoch_step = 1 if debug_training_enabled else 100
-    epoch_default = 1 if debug_training_enabled else 1000
 
-    gr.HTML("""
+    del dit_handler, llm_handler, init_params
+
+    epoch_min, epoch_step, epoch_default = _resolve_epoch_slider_defaults()
+
+    gr.HTML(
+        """
     <div style="text-align: center; padding: 10px; margin-bottom: 15px;">
         <h2>🎵 LoRA Training for ACE-Step</h2>
         <p>Build datasets from your audio files and train custom LoRA adapters</p>
     </div>
-    """)
-    
-    with gr.Tabs():
-        # ==================== Dataset Builder Tab ====================
-        with gr.Tab(t("training.tab_dataset_builder")):
-            create_help_button("training_dataset")
-            # ========== Load Existing OR Scan New ==========
-            gr.HTML(f"""
-            <div style="padding: 10px; margin-bottom: 10px; border: 1px solid #4a4a6a; border-radius: 8px; background: linear-gradient(135deg, #2a2a4a 0%, #1a1a3a 100%);">
-                <h3 style="margin: 0 0 5px 0;">{t("training.quick_start_title")}</h3>
-                <p style="margin: 0; color: #aaa;">Choose one: <b>Load existing dataset</b> OR <b>Scan new directory</b></p>
-            </div>
-            """)
-            
-            with gr.Row():
-                with gr.Column(scale=1):
-                    gr.HTML("<h4>📂 Load Existing Dataset</h4>")
-                    with gr.Row():
-                        load_json_path = gr.Textbox(
-                            label=t("training.load_dataset_label"),
-                            placeholder="./datasets/my_lora_dataset.json",
-                            info=t("training.load_dataset_info"), elem_classes=["has-info-container"],
-                            scale=3,
-                        )
-                        load_json_btn = gr.Button(t("training.load_btn"), variant="primary", scale=1)
-                    load_json_status = gr.Textbox(
-                        label=t("training.load_status"),
-                        interactive=False,
-                    )
-                
-                with gr.Column(scale=1):
-                    gr.HTML("<h4>🔍 Scan New Directory</h4>")
-                    with gr.Row():
-                        audio_directory = gr.Textbox(
-                            label=t("training.scan_label"),
-                            placeholder="/path/to/your/audio/folder",
-                            info=t("training.scan_info"), elem_classes=["has-info-container"],
-                            scale=3,
-                        )
-                        scan_btn = gr.Button(t("training.scan_btn"), variant="secondary", scale=1)
-                    scan_status = gr.Textbox(
-                        label=t("training.scan_status"),
-                        interactive=False,
-                    )
-            
-            gr.HTML("<hr>")
-            
-            with gr.Row():
-                with gr.Column(scale=2):
-                    
-                    # Audio files table
-                    audio_files_table = gr.Dataframe(
-                        headers=["#", "Filename", "Duration", "Lyrics", "Labeled", "BPM", "Key", "Caption"],
-                        datatype=["number", "str", "str", "str", "str", "str", "str", "str"],
-                        label=t("training.found_audio_files"),
-                        interactive=False,
-                        wrap=True,
-                    )
-                
-                with gr.Column(scale=1):
-                    gr.HTML(f"<h3>⚙️ {t('training.dataset_settings_header')}</h3>")
-                    
-                    dataset_name = gr.Textbox(
-                        label=t("training.dataset_name"),
-                        value="my_lora_dataset",
-                        placeholder=t("training.dataset_name_placeholder"),
-                    )
-                    
-                    all_instrumental = gr.Checkbox(
-                        label=t("training.all_instrumental"),
-                        value=True,
-                        info=t("training.all_instrumental_info"), elem_classes=["has-info-container"],
-                    )
-
-                    format_lyrics = gr.Checkbox(
-                        label="Format Lyrics (LM)",
-                        value=False,
-                        info="Use LM to format/structure user-provided lyrics from .txt files (coming soon)", elem_classes=["has-info-container"],
-                        interactive=False,  # Disabled for now - model update needed
-                    )
-
-                    transcribe_lyrics = gr.Checkbox(
-                        label="Transcribe Lyrics (LM)",
-                        value=False,
-                        info="Use LM to transcribe lyrics from audio (coming soon)", elem_classes=["has-info-container"],
-                        interactive=False,  # Disabled for now - model update needed
-                    )
-                    
-                    custom_tag = gr.Textbox(
-                        label=t("training.custom_tag"),
-                        placeholder="e.g., 8bit_retro, my_style",
-                        info=t("training.custom_tag_info"), elem_classes=["has-info-container"],
-                    )
-                    
-                    tag_position = gr.Radio(
-                        choices=[
-                            (t("training.tag_prepend"), "prepend"),
-                            (t("training.tag_append"), "append"),
-                            (t("training.tag_replace"), "replace"),
-                        ],
-                        value="replace",
-                        label=t("training.tag_position"),
-                        info=t("training.tag_position_info"), elem_classes=["has-info-container"],
-                    )
-
-                    genre_ratio = gr.Slider(
-                        minimum=0,
-                        maximum=100,
-                        step=10,
-                        value=0,
-                        label=t("training.genre_ratio"),
-                        info=t("training.genre_ratio_info"), elem_classes=["has-info-container"],
-                    )
-
-            gr.HTML(f"<hr><h3>🤖 {t('training.step2_title')}</h3>")
-            
-            with gr.Row():
-                with gr.Column(scale=3):
-                    gr.Markdown(t('training.step2_instruction'))
-                    skip_metas = gr.Checkbox(
-                        label=t("training.skip_metas"),
-                        value=False,
-                        info=t("training.skip_metas_info"), elem_classes=["has-info-container"],
-                    )
-                    only_unlabeled = gr.Checkbox(
-                        label=t("training.only_unlabeled"),
-                        value=False,
-                        info=t("training.only_unlabeled_info"), elem_classes=["has-info-container"],
-                    )
-                with gr.Column(scale=1):
-                    auto_label_btn = gr.Button(
-                        t("training.auto_label_btn"),
-                        variant="primary",
-                        size="lg",
-                    )
-            
-            label_progress = gr.Textbox(
-                label=t("training.label_progress"),
-                interactive=False,
-                lines=2,
-            )
-            
-            gr.HTML(f"<hr><h3>👀 {t('training.step3_title')}</h3>")
-            
-            with gr.Row():
-                with gr.Column(scale=1):
-                    sample_selector = gr.Slider(
-                        minimum=0,
-                        maximum=0,
-                        step=1,
-                        value=0,
-                        label=t("training.select_sample"),
-                        info=t("training.select_sample_info"), elem_classes=["has-info-container"],
-                    )
-                    
-                    preview_audio = gr.Audio(
-                        label=t("training.audio_preview"),
-                        type="filepath",
-                        interactive=False,
-                    )
-                    
-                    preview_filename = gr.Textbox(
-                        label=t("training.filename"),
-                        interactive=False,
-                    )
-
-                with gr.Column(scale=2):
-                    with gr.Row():
-                        edit_caption = gr.Textbox(
-                            label=t("training.caption"),
-                            lines=3,
-                            placeholder="Music description...",
-                        )
-
-                    with gr.Row():
-                        edit_genre = gr.Textbox(
-                            label=t("training.genre"),
-                            lines=1,
-                            placeholder="pop, electronic, dance...",
-                        )
-                        prompt_override = gr.Dropdown(
-                            choices=["Use Global Ratio", "Caption", "Genre"],
-                            value="Use Global Ratio",
-                            label=t("training.prompt_override_label"),
-                            info=t("training.prompt_override_info"), elem_classes=["has-info-container"],
-                        )
-
-                    with gr.Row():
-                        edit_lyrics = gr.Textbox(
-                            label=t("training.lyrics_editable_label"),
-                            lines=6,
-                            placeholder="[Verse 1]\nLyrics here...\n\n[Chorus]\n...",
-                        )
-                        raw_lyrics_display = gr.Textbox(
-                            label=t("training.raw_lyrics_label"),
-                            lines=6,
-                            placeholder=t("training.no_lyrics_placeholder"),
-                            interactive=False,  # Read-only, can copy but not edit
-                            visible=False,  # Hidden when no raw lyrics
-                        )
-                        has_raw_lyrics_state = gr.State(False)  # Track visibility
-
-                    with gr.Row():
-                        edit_bpm = gr.Number(
-                            label=t("training.bpm"),
-                            precision=0,
-                        )
-                        edit_keyscale = gr.Textbox(
-                            label=t("training.key_label"),
-                            placeholder=t("training.key_placeholder"),
-                        )
-                        edit_timesig = gr.Dropdown(
-                            choices=["", "2", "3", "4", "6", "N/A"],
-                            label=t("training.time_sig"),
-                        )
-                        edit_duration = gr.Number(
-                            label=t("training.duration_s"),
-                            precision=1,
-                            interactive=False,
-                        )
-                    
-                    with gr.Row():
-                        edit_language = gr.Dropdown(
-                            choices=["instrumental", "en", "zh", "ja", "ko", "es", "fr", "de", "pt", "ru", "unknown"],
-                            value="instrumental",
-                            label=t("training.language"),
-                        )
-                        edit_instrumental = gr.Checkbox(
-                            label=t("training.instrumental"),
-                            value=True,
-                        )
-                        save_edit_btn = gr.Button(t("training.save_changes_btn"), variant="secondary")
-                    
-                    edit_status = gr.Textbox(
-                        label=t("training.edit_status"),
-                        interactive=False,
-                    )
-            
-            gr.HTML(f"<hr><h3>💾 {t('training.step4_title')}</h3>")
-            
-            with gr.Row():
-                with gr.Column(scale=3):
-                    save_path = gr.Textbox(
-                        label=t("training.save_path"),
-                        value="./datasets/my_lora_dataset.json",
-                        placeholder="./datasets/dataset_name.json",
-                        info=t("training.save_path_info"), elem_classes=["has-info-container"],
-                    )
-                with gr.Column(scale=1):
-                    save_dataset_btn = gr.Button(
-                        t("training.save_dataset_btn"),
-                        variant="primary",
-                        size="lg",
-                    )
-            
-            save_status = gr.Textbox(
-                label=t("training.save_status"),
-                interactive=False,
-                lines=2,
-            )
-            
-            gr.HTML(f"<hr><h3>⚡ {t('training.step5_title')}</h3>")
-            
-            gr.Markdown(t('training.step5_intro'))
-            
-            with gr.Row():
-                with gr.Column(scale=3):
-                    load_existing_dataset_path = gr.Textbox(
-                        label=t("training.load_existing_label"),
-                        placeholder="./datasets/my_lora_dataset.json",
-                        info=t("training.load_existing_info"), elem_classes=["has-info-container"],
-                    )
-                with gr.Column(scale=1):
-                    load_existing_dataset_btn = gr.Button(
-                        t("training.load_dataset_btn"),
-                        variant="secondary",
-                        size="lg",
-                    )
-            
-            load_existing_status = gr.Textbox(
-                label=t("training.load_status"),
-                interactive=False,
-            )
-            
-            gr.Markdown(t('training.step5_details'))
-
-            with gr.Row():
-                preprocess_mode = gr.Dropdown(
-                    label="Preprocess For",
-                    choices=["LoRA", "LoKr"],
-                    value="LoRA",
-                    info="LoRA keeps compatibility mode; LoKr uses per-sample source-style context.", elem_classes=["has-info-container"],
-                )
-
-            with gr.Row():
-                with gr.Column(scale=3):
-                    preprocess_output_dir = gr.Textbox(
-                        label=t("training.tensor_output_dir"),
-                        value="./datasets/preprocessed_tensors",
-                        placeholder="./datasets/preprocessed_tensors",
-                        info=t("training.tensor_output_info"), elem_classes=["has-info-container"],
-                    )
-                with gr.Column(scale=1):
-                    preprocess_btn = gr.Button(
-                        t("training.preprocess_btn"),
-                        variant="primary",
-                        size="lg",
-                    )
-            
-            preprocess_progress = gr.Textbox(
-                label=t("training.preprocess_progress"),
-                interactive=False,
-                lines=3,
-            )
-        
-        # ==================== Training Tab ====================
-        with gr.Tab(t("training.tab_train_lora")):
-            create_help_button("training_train")
-            with gr.Row():
-                with gr.Column(scale=2):
-                    gr.HTML(f"<h3>📊 {t('training.train_section_tensors')}</h3>")
-                    
-                    gr.Markdown(t('training.train_tensor_selection_desc'))
-                    
-                    training_tensor_dir = gr.Textbox(
-                        label=t("training.preprocessed_tensors_dir"),
-                        placeholder="./datasets/preprocessed_tensors",
-                        value="./datasets/preprocessed_tensors",
-                        info=t("training.preprocessed_tensors_info"), elem_classes=["has-info-container"],
-                    )
-                    
-                    load_dataset_btn = gr.Button(t("training.load_dataset_btn"), variant="secondary")
-                    
-                    training_dataset_info = gr.Textbox(
-                        label=t("training.dataset_info"),
-                        interactive=False,
-                        lines=3,
-                    )
-                
-                with gr.Column(scale=1):
-                    gr.HTML(f"<h3>⚙️ {t('training.train_section_lora')}</h3>")
-                    
-                    lora_rank = gr.Slider(
-                        minimum=4,
-                        maximum=256,
-                        step=4,
-                        value=64,
-                        label=t("training.lora_rank"),
-                        info=t("training.lora_rank_info"), elem_classes=["has-info-container"],
-                    )
-                    
-                    lora_alpha = gr.Slider(
-                        minimum=4,
-                        maximum=512,
-                        step=4,
-                        value=128,
-                        label=t("training.lora_alpha"),
-                        info=t("training.lora_alpha_info"), elem_classes=["has-info-container"],
-                    )
-                    
-                    lora_dropout = gr.Slider(
-                        minimum=0.0,
-                        maximum=0.5,
-                        step=0.05,
-                        value=0.1,
-                        label=t("training.lora_dropout"),
-                    )
-            
-            gr.HTML(f"<hr><h3>🎛️ {t('training.train_section_params')}</h3>")
-            
-            with gr.Row():
-                learning_rate = gr.Number(
-                    label=t("training.learning_rate"),
-                    value=3e-4,
-                    info=t("training.learning_rate_info"), elem_classes=["has-info-container"],
-                )
-                
-                train_epochs = gr.Slider(
-                    minimum=epoch_min,
-                    maximum=4000,
-                    step=epoch_step,
-                    value=epoch_default,
-                    label=t("training.max_epochs"),
-                )
-                
-                train_batch_size = gr.Slider(
-                    minimum=1,
-                    maximum=8,
-                    step=1,
-                    value=1,
-                    label=t("training.batch_size"),
-                    info=t("training.batch_size_info"), elem_classes=["has-info-container"],
-                )
-                
-                gradient_accumulation = gr.Slider(
-                    minimum=1,
-                    maximum=16,
-                    step=1,
-                    value=1,
-                    label=t("training.gradient_accumulation"),
-                    info=t("training.gradient_accumulation_info"), elem_classes=["has-info-container"],
-                )
-            
-            with gr.Row():
-                save_every_n_epochs = gr.Slider(
-                    minimum=1,
-                    maximum=1000,
-                    step=1,
-                    value=10,
-                    label=t("training.save_every_n_epochs"),
-                )
-                
-                training_shift = gr.Slider(
-                    minimum=1.0,
-                    maximum=5.0,
-                    step=0.5,
-                    value=3.0,
-                    label=t("training.shift"),
-                    info=t("training.shift_info"), elem_classes=["has-info-container"],
-                )
-                
-                training_seed = gr.Number(
-                    label=t("training.seed"),
-                    value=42,
-                    precision=0,
-                )
-            
-            with gr.Row():
-                lora_output_dir = gr.Textbox(
-                    label=t("training.output_dir"),
-                    value="./lora_output",
-                    placeholder="./lora_output",
-                    info=t("training.output_dir_info"), elem_classes=["has-info-container"],
-                )
-            
-            with gr.Row():
-                resume_checkpoint_dir = gr.Textbox(
-                    label="Resume Checkpoint",
-                    placeholder="./lora_output/checkpoints/epoch_200",
-                    info="Directory of a saved LoRA checkpoint to resume from", elem_classes=["has-info-container"],
-                )
-            
-            gr.HTML("<hr>")
-            
-            with gr.Row():
-                with gr.Column(scale=1):
-                    start_training_btn = gr.Button(
-                        t("training.start_training_btn"),
-                        variant="primary",
-                        size="lg",
-                    )
-                with gr.Column(scale=1):
-                    stop_training_btn = gr.Button(
-                        t("training.stop_training_btn"),
-                        variant="stop",
-                        size="lg",
-                    )
-            
-            training_progress = gr.Textbox(
-                label=t("training.training_progress"),
-                interactive=False,
-                lines=2,
-            )
-            
-            with gr.Row():
-                training_log = gr.Textbox(
-                    label=t("training.training_log"),
-                    interactive=False,
-                    lines=10,
-                    max_lines=15,
-                    scale=1,
-                )
-                training_loss_plot = gr.Plot(
-                    label=t("training.training_loss_title"),
-                    scale=1,
-                )
-            
-            gr.HTML(f"<hr><h3>📦 {t('training.export_header')}</h3>")
-            
-            with gr.Row():
-                export_path = gr.Textbox(
-                    label=t("training.export_path"),
-                    value="./lora_output/final_lora",
-                    placeholder="./lora_output/my_lora",
-                )
-                export_lora_btn = gr.Button(t("training.export_lora_btn"), variant="secondary")
-            
-            export_status = gr.Textbox(
-                label=t("training.export_status"),
-                interactive=False,
-            )
-
-        # ==================== Train LoKr Tab ====================
-        with gr.Tab("🚀 Train LoKr"):
-            create_help_button("training_lokr")
-            with gr.Row():
-                with gr.Column(scale=2):
-                    gr.HTML("<h3>📊 Preprocessed Tensors</h3>")
-                    gr.Markdown(
-                        "Select the directory containing preprocessed tensor files (`.pt` files). "
-                        "These are created in the Dataset Builder tab."
-                    )
-
-                    lokr_training_tensor_dir = gr.Textbox(
-                        label="Preprocessed Tensors Directory",
-                        placeholder="./datasets/preprocessed_tensors",
-                        value="./datasets/preprocessed_tensors",
-                        info="Path to directory containing manifest.json and tensor .pt files.", elem_classes=["has-info-container"],
-                    )
-
-                    lokr_load_dataset_btn = gr.Button("Load Dataset", variant="secondary")
-
-                    lokr_training_dataset_info = gr.Textbox(
-                        label="Dataset Info",
-                        interactive=False,
-                        lines=3,
-                    )
-
-                with gr.Column(scale=1):
-                    gr.HTML("<h3>⚙️ LoKr Settings</h3>")
-
-                    lokr_linear_dim = gr.Slider(
-                        minimum=4,
-                        maximum=256,
-                        step=4,
-                        value=64,
-                        label="LoKr Linear Dim",
-                        info="Adapter rank-like width for LoKr linear layers.", elem_classes=["has-info-container"],
-                    )
-                    lokr_linear_alpha = gr.Slider(
-                        minimum=4,
-                        maximum=512,
-                        step=4,
-                        value=128,
-                        label="LoKr Linear Alpha",
-                        info="Scaling factor for LoKr adapters.", elem_classes=["has-info-container"],
-                    )
-                    lokr_factor = gr.Number(
-                        label="LoKr Factor",
-                        value=-1,
-                        precision=0,
-                        info="-1 uses automatic Kronecker factor selection.", elem_classes=["has-info-container"],
-                    )
-                    lokr_decompose_both = gr.Checkbox(
-                        label="Decompose Both",
-                        value=False,
-                        info="Enable decomposition on both matrices.", elem_classes=["has-info-container"],
-                    )
-                    lokr_use_tucker = gr.Checkbox(
-                        label="Use Tucker",
-                        value=False,
-                        info="Enable Tucker decomposition mode.", elem_classes=["has-info-container"],
-                    )
-                    lokr_use_scalar = gr.Checkbox(
-                        label="Use Scalar",
-                        value=False,
-                        info="Enable scalar calibration in LyCORIS.", elem_classes=["has-info-container"],
-                    )
-                    lokr_weight_decompose = gr.Checkbox(
-                        label="Weight Decompose (DoRA)",
-                        value=True,
-                        info="Enable DoRA-style weight decomposition when supported.", elem_classes=["has-info-container"],
-                    )
-
-            gr.HTML("<hr><h3>🎛️ Training Parameters</h3>")
-
-            with gr.Row():
-                lokr_learning_rate = gr.Number(
-                    label="Learning Rate",
-                    value=1e-3,
-                    info="LoKr commonly uses a higher LR than LoRA. Tune per dataset.", elem_classes=["has-info-container"],
-                )
-
-                lokr_train_epochs = gr.Slider(
-                    minimum=1,
-                    maximum=4000,
-                    step=1,
-                    value=500,
-                    label="Max Epochs",
-                )
-
-                lokr_train_batch_size = gr.Slider(
-                    minimum=1,
-                    maximum=8,
-                    step=1,
-                    value=1,
-                    label="Batch Size",
-                )
-
-                lokr_gradient_accumulation = gr.Slider(
-                    minimum=1,
-                    maximum=16,
-                    step=1,
-                    value=4,
-                    label="Gradient Accumulation",
-                )
-
-            with gr.Row():
-                lokr_save_every_n_epochs = gr.Slider(
-                    minimum=1,
-                    maximum=1000,
-                    step=1,
-                    value=10,
-                    label="Save Every N Epochs",
-                )
-
-                lokr_training_shift = gr.Slider(
-                    minimum=1.0,
-                    maximum=5.0,
-                    step=0.5,
-                    value=3.0,
-                    label="Shift",
-                    info="Turbo model training timestep shift.", elem_classes=["has-info-container"],
-                )
-
-                lokr_training_seed = gr.Number(
-                    label="Seed",
-                    value=42,
-                    precision=0,
-                )
-
-            with gr.Row():
-                lokr_output_dir = gr.Textbox(
-                    label="Output Directory",
-                    value="./lokr_output",
-                    placeholder="./lokr_output",
-                    info="Where LoKr checkpoints and final weights will be written.", elem_classes=["has-info-container"],
-                )
-
-            gr.HTML("<hr>")
-
-            with gr.Row():
-                with gr.Column(scale=1):
-                    start_lokr_training_btn = gr.Button(
-                        "Start LoKr Training",
-                        variant="primary",
-                        size="lg",
-                    )
-                with gr.Column(scale=1):
-                    stop_lokr_training_btn = gr.Button(
-                        "Stop Training",
-                        variant="stop",
-                        size="lg",
-                    )
-
-            lokr_training_progress = gr.Textbox(
-                label="Training Progress",
-                interactive=False,
-                lines=2,
-            )
-
-            with gr.Row():
-                lokr_training_log = gr.Textbox(
-                    label="Training Log",
-                    interactive=False,
-                    lines=10,
-                    max_lines=15,
-                    scale=1,
-                )
-                lokr_training_loss_plot = gr.Plot(
-                    label="LoKr Training Loss",
-                    scale=1,
-                )
-
-            gr.HTML("<hr><h3>📦 Export LoKr</h3>")
-
-            with gr.Row():
-                lokr_export_path = gr.Textbox(
-                    label="Export Path",
-                    value="./lokr_output/final_lokr",
-                    placeholder="./lokr_output/my_lokr",
-                )
-                export_lokr_btn = gr.Button("📦 Export LoKr", variant="secondary")
-
-            with gr.Row():
-                lokr_export_epoch = gr.Dropdown(
-                    choices=["Latest (auto)"],
-                    value="Latest (auto)",
-                    label="Checkpoint Epoch",
-                    info="Select a specific epoch checkpoint to export, or keep Latest (auto).", elem_classes=["has-info-container"],
-                )
-                refresh_lokr_export_epochs_btn = gr.Button("↻ Refresh Epochs", variant="secondary")
+    """
+    )
 
-            lokr_export_status = gr.Textbox(
-                label="Export Status",
-                interactive=False,
+    training_section: dict[str, object] = {}
+    with gr.Tabs():
+        training_section.update(create_dataset_builder_tab())
+        training_section.update(
+            create_training_lora_tab(
+                epoch_min=epoch_min,
+                epoch_step=epoch_step,
+                epoch_default=epoch_default,
             )
-
-    # Store dataset builder state
-    dataset_builder_state = gr.State(None)
-    training_state = gr.State({"is_training": False, "should_stop": False})
-    
-    return {
-        # Dataset Builder - Load or Scan
-        "load_json_path": load_json_path,
-        "load_json_btn": load_json_btn,
-        "load_json_status": load_json_status,
-        "audio_directory": audio_directory,
-        "scan_btn": scan_btn,
-        "scan_status": scan_status,
-        "audio_files_table": audio_files_table,
-        "dataset_name": dataset_name,
-        "all_instrumental": all_instrumental,
-        "format_lyrics": format_lyrics,
-        "transcribe_lyrics": transcribe_lyrics,
-        "custom_tag": custom_tag,
-        "tag_position": tag_position,
-        "skip_metas": skip_metas,
-        "only_unlabeled": only_unlabeled,
-        "auto_label_btn": auto_label_btn,
-        "label_progress": label_progress,
-        "sample_selector": sample_selector,
-        "preview_audio": preview_audio,
-        "preview_filename": preview_filename,
-        "edit_caption": edit_caption,
-        "edit_genre": edit_genre,
-        "prompt_override": prompt_override,
-        "genre_ratio": genre_ratio,
-        "edit_lyrics": edit_lyrics,
-        "raw_lyrics_display": raw_lyrics_display,
-        "has_raw_lyrics_state": has_raw_lyrics_state,
-        "edit_bpm": edit_bpm,
-        "edit_keyscale": edit_keyscale,
-        "edit_timesig": edit_timesig,
-        "edit_duration": edit_duration,
-        "edit_language": edit_language,
-        "edit_instrumental": edit_instrumental,
-        "save_edit_btn": save_edit_btn,
-        "edit_status": edit_status,
-        "save_path": save_path,
-        "save_dataset_btn": save_dataset_btn,
-        "save_status": save_status,
-        # Preprocessing
-        "load_existing_dataset_path": load_existing_dataset_path,
-        "load_existing_dataset_btn": load_existing_dataset_btn,
-        "load_existing_status": load_existing_status,
-        "preprocess_mode": preprocess_mode,
-        "preprocess_output_dir": preprocess_output_dir,
-        "preprocess_btn": preprocess_btn,
-        "preprocess_progress": preprocess_progress,
-        "dataset_builder_state": dataset_builder_state,
-        # Training
-        "training_tensor_dir": training_tensor_dir,
-        "load_dataset_btn": load_dataset_btn,
-        "training_dataset_info": training_dataset_info,
-        "lora_rank": lora_rank,
-        "lora_alpha": lora_alpha,
-        "lora_dropout": lora_dropout,
-        "learning_rate": learning_rate,
-        "train_epochs": train_epochs,
-        "train_batch_size": train_batch_size,
-        "gradient_accumulation": gradient_accumulation,
-        "save_every_n_epochs": save_every_n_epochs,
-        "training_shift": training_shift,
-        "training_seed": training_seed,
-        "lora_output_dir": lora_output_dir,
-        "resume_checkpoint_dir": resume_checkpoint_dir,
-        "start_training_btn": start_training_btn,
-        "stop_training_btn": stop_training_btn,
-        "training_progress": training_progress,
-        "training_log": training_log,
-        "training_loss_plot": training_loss_plot,
-        "export_path": export_path,
-        "export_lora_btn": export_lora_btn,
-        "export_status": export_status,
-        # LoKr training
-        "lokr_training_tensor_dir": lokr_training_tensor_dir,
-        "lokr_load_dataset_btn": lokr_load_dataset_btn,
-        "lokr_training_dataset_info": lokr_training_dataset_info,
-        "lokr_linear_dim": lokr_linear_dim,
-        "lokr_linear_alpha": lokr_linear_alpha,
-        "lokr_factor": lokr_factor,
-        "lokr_decompose_both": lokr_decompose_both,
-        "lokr_use_tucker": lokr_use_tucker,
-        "lokr_use_scalar": lokr_use_scalar,
-        "lokr_weight_decompose": lokr_weight_decompose,
-        "lokr_learning_rate": lokr_learning_rate,
-        "lokr_train_epochs": lokr_train_epochs,
-        "lokr_train_batch_size": lokr_train_batch_size,
-        "lokr_gradient_accumulation": lokr_gradient_accumulation,
-        "lokr_save_every_n_epochs": lokr_save_every_n_epochs,
-        "lokr_training_shift": lokr_training_shift,
-        "lokr_training_seed": lokr_training_seed,
-        "lokr_output_dir": lokr_output_dir,
-        "start_lokr_training_btn": start_lokr_training_btn,
-        "stop_lokr_training_btn": stop_lokr_training_btn,
-        "lokr_training_progress": lokr_training_progress,
-        "lokr_training_log": lokr_training_log,
-        "lokr_training_loss_plot": lokr_training_loss_plot,
-        "lokr_export_path": lokr_export_path,
-        "lokr_export_epoch": lokr_export_epoch,
-        "refresh_lokr_export_epochs_btn": refresh_lokr_export_epochs_btn,
-        "export_lokr_btn": export_lokr_btn,
-        "lokr_export_status": lokr_export_status,
-        "training_state": training_state,
-    }
+        )
+        training_section.update(create_training_lokr_tab())
+        dataset_builder_state = gr.State(None)
+        training_state = gr.State({"is_training": False, "should_stop": False})
+        training_section.update(
+            {
+                "dataset_builder_state": dataset_builder_state,
+                "training_state": training_state,
+            }
+        )
+    return training_section
diff --git a/acestep/ui/gradio/interfaces/training_contract_ast_utils.py b/acestep/ui/gradio/interfaces/training_contract_ast_utils.py
new file mode 100644
index 00000000..967c807a
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_contract_ast_utils.py
@@ -0,0 +1,77 @@
+"""AST utility helpers for training interface decomposition contract tests."""
+
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+
+INTERFACES_DIR = Path(__file__).resolve().parent
+WIRING_DIR = INTERFACES_DIR.parent / "events" / "wiring"
+
+
+def load_module(module_name: str) -> ast.Module:
+    """Parse and return AST for an interfaces module."""
+
+    path = INTERFACES_DIR / module_name
+    return ast.parse(path.read_text(encoding="utf-8"))
+
+
+def call_name(node: ast.AST) -> str | None:
+    """Extract a simple call-target name from an AST call function node."""
+
+    if isinstance(node, ast.Name):
+        return node.id
+    if isinstance(node, ast.Attribute):
+        return node.attr
+    return None
+
+
+def collect_return_dict_keys(module_name: str, function_name: str) -> set[str]:
+    """Collect string keys from dict literals assigned/updated/returned in a function."""
+
+    module = load_module(module_name)
+    function_node = None
+    for node in module.body:
+        if isinstance(node, ast.FunctionDef) and node.name == function_name:
+            function_node = node
+            break
+    if function_node is None:
+        raise AssertionError(f"{function_name} not found in {module_name}")
+
+    keys: set[str] = set()
+    for node in ast.walk(function_node):
+        if isinstance(node, ast.Assign) and isinstance(node.value, ast.Dict):
+            for key in node.value.keys:
+                if isinstance(key, ast.Constant) and isinstance(key.value, str):
+                    keys.add(key.value)
+        if isinstance(node, ast.AnnAssign) and isinstance(node.value, ast.Dict):
+            for key in node.value.keys:
+                if isinstance(key, ast.Constant) and isinstance(key.value, str):
+                    keys.add(key.value)
+        if isinstance(node, ast.Call) and isinstance(node.func, ast.Attribute) and node.func.attr == "update":
+            if node.args and isinstance(node.args[0], ast.Dict):
+                for key in node.args[0].keys:
+                    if isinstance(key, ast.Constant) and isinstance(key.value, str):
+                        keys.add(key.value)
+        if isinstance(node, ast.Return) and isinstance(node.value, ast.Dict):
+            for key in node.value.keys:
+                if isinstance(key, ast.Constant) and isinstance(key.value, str):
+                    keys.add(key.value)
+    return keys
+
+
+def collect_training_section_keys_used_by_wiring() -> set[str]:
+    """Collect ``training_section[...]`` keys referenced by training wiring modules."""
+
+    keys: set[str] = set()
+    for path in WIRING_DIR.glob("training_*wiring.py"):
+        module = ast.parse(path.read_text(encoding="utf-8"))
+        for node in ast.walk(module):
+            if not isinstance(node, ast.Subscript):
+                continue
+            if not isinstance(node.value, ast.Name) or node.value.id != "training_section":
+                continue
+            if isinstance(node.slice, ast.Constant) and isinstance(node.slice.value, str):
+                keys.add(node.slice.value)
+    return keys
diff --git a/acestep/ui/gradio/interfaces/training_dataset_builder_tab.py b/acestep/ui/gradio/interfaces/training_dataset_builder_tab.py
new file mode 100644
index 00000000..87551c0a
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_dataset_builder_tab.py
@@ -0,0 +1,29 @@
+"""Dataset-builder tab facade for the Gradio training interface."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.help_content import create_help_button
+from acestep.ui.gradio.i18n import t
+from acestep.ui.gradio.interfaces.training_dataset_tab_label_preview import (
+    build_dataset_label_and_preview_controls,
+)
+from acestep.ui.gradio.interfaces.training_dataset_tab_save_preprocess import (
+    build_dataset_save_and_preprocess_controls,
+)
+from acestep.ui.gradio.interfaces.training_dataset_tab_scan_settings import (
+    build_dataset_scan_and_settings_controls,
+)
+
+
+def create_dataset_builder_tab() -> dict[str, object]:
+    """Create the Dataset Builder tab and return all exposed component handles."""
+
+    with gr.Tab(t("training.tab_dataset_builder")):
+        create_help_button("training_dataset")
+        tab_controls: dict[str, object] = {}
+        tab_controls.update(build_dataset_scan_and_settings_controls())
+        tab_controls.update(build_dataset_label_and_preview_controls())
+        tab_controls.update(build_dataset_save_and_preprocess_controls())
+    return tab_controls
diff --git a/acestep/ui/gradio/interfaces/training_dataset_tab_label_preview.py b/acestep/ui/gradio/interfaces/training_dataset_tab_label_preview.py
new file mode 100644
index 00000000..dcc4838f
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_dataset_tab_label_preview.py
@@ -0,0 +1,175 @@
+"""Dataset labeling and sample-preview controls for the training dataset tab."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_dataset_label_and_preview_controls() -> dict[str, object]:
+    """Render auto-label and sample-preview editors for dataset-builder workflows."""
+
+    gr.HTML(f"<hr><h3>🤖 {t('training.step2_title')}</h3>")
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            gr.Markdown(t("training.step2_instruction"))
+            skip_metas = gr.Checkbox(
+                label=t("training.skip_metas"),
+                value=False,
+                info=t("training.skip_metas_info"),
+                elem_classes=["has-info-container"],
+            )
+            only_unlabeled = gr.Checkbox(
+                label=t("training.only_unlabeled"),
+                value=False,
+                info=t("training.only_unlabeled_info"),
+                elem_classes=["has-info-container"],
+            )
+        with gr.Column(scale=1):
+            auto_label_btn = gr.Button(
+                t("training.auto_label_btn"),
+                variant="primary",
+                size="lg",
+            )
+
+    label_progress = gr.Textbox(
+        label=t("training.label_progress"),
+        interactive=False,
+        lines=2,
+    )
+
+    gr.HTML(f"<hr><h3>👀 {t('training.step3_title')}</h3>")
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            sample_selector = gr.Slider(
+                minimum=0,
+                maximum=0,
+                step=1,
+                value=0,
+                label=t("training.select_sample"),
+                info=t("training.select_sample_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            preview_audio = gr.Audio(
+                label=t("training.audio_preview"),
+                type="filepath",
+                interactive=False,
+            )
+
+            preview_filename = gr.Textbox(
+                label=t("training.filename"),
+                interactive=False,
+            )
+
+        with gr.Column(scale=2):
+            with gr.Row():
+                edit_caption = gr.Textbox(
+                    label=t("training.caption"),
+                    lines=3,
+                    placeholder="Music description...",
+                )
+
+            with gr.Row():
+                edit_genre = gr.Textbox(
+                    label=t("training.genre"),
+                    lines=1,
+                    placeholder="pop, electronic, dance...",
+                )
+                prompt_override = gr.Dropdown(
+                    choices=["Use Global Ratio", "Caption", "Genre"],
+                    value="Use Global Ratio",
+                    label=t("training.prompt_override_label"),
+                    info=t("training.prompt_override_info"),
+                    elem_classes=["has-info-container"],
+                )
+
+            with gr.Row():
+                edit_lyrics = gr.Textbox(
+                    label=t("training.lyrics_editable_label"),
+                    lines=6,
+                    placeholder="[Verse 1]\nLyrics here...\n\n[Chorus]\n...",
+                )
+                raw_lyrics_display = gr.Textbox(
+                    label=t("training.raw_lyrics_label"),
+                    lines=6,
+                    placeholder=t("training.no_lyrics_placeholder"),
+                    interactive=False,
+                    visible=False,
+                )
+                has_raw_lyrics_state = gr.State(False)
+
+            with gr.Row():
+                edit_bpm = gr.Number(
+                    label=t("training.bpm"),
+                    precision=0,
+                )
+                edit_keyscale = gr.Textbox(
+                    label=t("training.key_label"),
+                    placeholder=t("training.key_placeholder"),
+                )
+                edit_timesig = gr.Dropdown(
+                    choices=["", "2", "3", "4", "6", "N/A"],
+                    label=t("training.time_sig"),
+                )
+                edit_duration = gr.Number(
+                    label=t("training.duration_s"),
+                    precision=1,
+                    interactive=False,
+                )
+
+            with gr.Row():
+                edit_language = gr.Dropdown(
+                    choices=[
+                        "instrumental",
+                        "en",
+                        "zh",
+                        "ja",
+                        "ko",
+                        "es",
+                        "fr",
+                        "de",
+                        "pt",
+                        "ru",
+                        "unknown",
+                    ],
+                    value="instrumental",
+                    label=t("training.language"),
+                )
+                edit_instrumental = gr.Checkbox(
+                    label=t("training.instrumental"),
+                    value=True,
+                )
+                save_edit_btn = gr.Button(t("training.save_changes_btn"), variant="secondary")
+
+            edit_status = gr.Textbox(
+                label=t("training.edit_status"),
+                interactive=False,
+            )
+
+    return {
+        "skip_metas": skip_metas,
+        "only_unlabeled": only_unlabeled,
+        "auto_label_btn": auto_label_btn,
+        "label_progress": label_progress,
+        "sample_selector": sample_selector,
+        "preview_audio": preview_audio,
+        "preview_filename": preview_filename,
+        "edit_caption": edit_caption,
+        "edit_genre": edit_genre,
+        "prompt_override": prompt_override,
+        "edit_lyrics": edit_lyrics,
+        "raw_lyrics_display": raw_lyrics_display,
+        "has_raw_lyrics_state": has_raw_lyrics_state,
+        "edit_bpm": edit_bpm,
+        "edit_keyscale": edit_keyscale,
+        "edit_timesig": edit_timesig,
+        "edit_duration": edit_duration,
+        "edit_language": edit_language,
+        "edit_instrumental": edit_instrumental,
+        "save_edit_btn": save_edit_btn,
+        "edit_status": edit_status,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_dataset_tab_save_preprocess.py b/acestep/ui/gradio/interfaces/training_dataset_tab_save_preprocess.py
new file mode 100644
index 00000000..2729ae0a
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_dataset_tab_save_preprocess.py
@@ -0,0 +1,105 @@
+"""Dataset save and preprocess controls for the training dataset tab."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_dataset_save_and_preprocess_controls() -> dict[str, object]:
+    """Render dataset save/load-preprocess controls and return component handles."""
+
+    gr.HTML(f"<hr><h3>💾 {t('training.step4_title')}</h3>")
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            save_path = gr.Textbox(
+                label=t("training.save_path"),
+                value="./datasets/my_lora_dataset.json",
+                placeholder="./datasets/dataset_name.json",
+                info=t("training.save_path_info"),
+                elem_classes=["has-info-container"],
+            )
+        with gr.Column(scale=1):
+            save_dataset_btn = gr.Button(
+                t("training.save_dataset_btn"),
+                variant="primary",
+                size="lg",
+            )
+
+    save_status = gr.Textbox(
+        label=t("training.save_status"),
+        interactive=False,
+        lines=2,
+    )
+
+    gr.HTML(f"<hr><h3>⚡ {t('training.step5_title')}</h3>")
+
+    gr.Markdown(t("training.step5_intro"))
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            load_existing_dataset_path = gr.Textbox(
+                label=t("training.load_existing_label"),
+                placeholder="./datasets/my_lora_dataset.json",
+                info=t("training.load_existing_info"),
+                elem_classes=["has-info-container"],
+            )
+        with gr.Column(scale=1):
+            load_existing_dataset_btn = gr.Button(
+                t("training.load_dataset_btn"),
+                variant="secondary",
+                size="lg",
+            )
+
+    load_existing_status = gr.Textbox(
+        label=t("training.load_status"),
+        interactive=False,
+    )
+
+    gr.Markdown(t("training.step5_details"))
+
+    with gr.Row():
+        preprocess_mode = gr.Dropdown(
+            label="Preprocess For",
+            choices=["LoRA", "LoKr"],
+            value="LoRA",
+            info="LoRA keeps compatibility mode; LoKr uses per-sample source-style context.",
+            elem_classes=["has-info-container"],
+        )
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            preprocess_output_dir = gr.Textbox(
+                label=t("training.tensor_output_dir"),
+                value="./datasets/preprocessed_tensors",
+                placeholder="./datasets/preprocessed_tensors",
+                info=t("training.tensor_output_info"),
+                elem_classes=["has-info-container"],
+            )
+        with gr.Column(scale=1):
+            preprocess_btn = gr.Button(
+                t("training.preprocess_btn"),
+                variant="primary",
+                size="lg",
+            )
+
+    preprocess_progress = gr.Textbox(
+        label=t("training.preprocess_progress"),
+        interactive=False,
+        lines=3,
+    )
+
+    return {
+        "save_path": save_path,
+        "save_dataset_btn": save_dataset_btn,
+        "save_status": save_status,
+        "load_existing_dataset_path": load_existing_dataset_path,
+        "load_existing_dataset_btn": load_existing_dataset_btn,
+        "load_existing_status": load_existing_status,
+        "preprocess_mode": preprocess_mode,
+        "preprocess_output_dir": preprocess_output_dir,
+        "preprocess_btn": preprocess_btn,
+        "preprocess_progress": preprocess_progress,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_dataset_tab_scan_settings.py b/acestep/ui/gradio/interfaces/training_dataset_tab_scan_settings.py
new file mode 100644
index 00000000..b7b7a463
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_dataset_tab_scan_settings.py
@@ -0,0 +1,143 @@
+"""Dataset scan/load and settings controls for the training dataset tab."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_dataset_scan_and_settings_controls() -> dict[str, object]:
+    """Render scan/load controls and dataset settings for the dataset-builder tab."""
+
+    gr.HTML(
+        f"""
+            <div style="padding: 10px; margin-bottom: 10px; border: 1px solid #4a4a6a; border-radius: 8px; background: linear-gradient(135deg, #2a2a4a 0%, #1a1a3a 100%);">
+                <h3 style="margin: 0 0 5px 0;">{t("training.quick_start_title")}</h3>
+                <p style="margin: 0; color: #aaa;">Choose one: <b>Load existing dataset</b> OR <b>Scan new directory</b></p>
+            </div>
+            """
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.HTML("<h4>📂 Load Existing Dataset</h4>")
+            with gr.Row():
+                load_json_path = gr.Textbox(
+                    label=t("training.load_dataset_label"),
+                    placeholder="./datasets/my_lora_dataset.json",
+                    info=t("training.load_dataset_info"),
+                    elem_classes=["has-info-container"],
+                    scale=3,
+                )
+                load_json_btn = gr.Button(t("training.load_btn"), variant="primary", scale=1)
+            load_json_status = gr.Textbox(
+                label=t("training.load_status"),
+                interactive=False,
+            )
+
+        with gr.Column(scale=1):
+            gr.HTML("<h4>🔍 Scan New Directory</h4>")
+            with gr.Row():
+                audio_directory = gr.Textbox(
+                    label=t("training.scan_label"),
+                    placeholder="/path/to/your/audio/folder",
+                    info=t("training.scan_info"),
+                    elem_classes=["has-info-container"],
+                    scale=3,
+                )
+                scan_btn = gr.Button(t("training.scan_btn"), variant="secondary", scale=1)
+            scan_status = gr.Textbox(
+                label=t("training.scan_status"),
+                interactive=False,
+            )
+
+    gr.HTML("<hr>")
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            audio_files_table = gr.Dataframe(
+                headers=["#", "Filename", "Duration", "Lyrics", "Labeled", "BPM", "Key", "Caption"],
+                datatype=["number", "str", "str", "str", "str", "str", "str", "str"],
+                label=t("training.found_audio_files"),
+                interactive=False,
+                wrap=True,
+            )
+
+        with gr.Column(scale=1):
+            gr.HTML(f"<h3>⚙️ {t('training.dataset_settings_header')}</h3>")
+
+            dataset_name = gr.Textbox(
+                label=t("training.dataset_name"),
+                value="my_lora_dataset",
+                placeholder=t("training.dataset_name_placeholder"),
+            )
+
+            all_instrumental = gr.Checkbox(
+                label=t("training.all_instrumental"),
+                value=True,
+                info=t("training.all_instrumental_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            format_lyrics = gr.Checkbox(
+                label="Format Lyrics (LM)",
+                value=False,
+                info="Use LM to format/structure user-provided lyrics from .txt files (coming soon)",
+                elem_classes=["has-info-container"],
+                interactive=False,
+            )
+
+            transcribe_lyrics = gr.Checkbox(
+                label="Transcribe Lyrics (LM)",
+                value=False,
+                info="Use LM to transcribe lyrics from audio (coming soon)",
+                elem_classes=["has-info-container"],
+                interactive=False,
+            )
+
+            custom_tag = gr.Textbox(
+                label=t("training.custom_tag"),
+                placeholder="e.g., 8bit_retro, my_style",
+                info=t("training.custom_tag_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            tag_position = gr.Radio(
+                choices=[
+                    (t("training.tag_prepend"), "prepend"),
+                    (t("training.tag_append"), "append"),
+                    (t("training.tag_replace"), "replace"),
+                ],
+                value="replace",
+                label=t("training.tag_position"),
+                info=t("training.tag_position_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            genre_ratio = gr.Slider(
+                minimum=0,
+                maximum=100,
+                step=10,
+                value=0,
+                label=t("training.genre_ratio"),
+                info=t("training.genre_ratio_info"),
+                elem_classes=["has-info-container"],
+            )
+
+    return {
+        "load_json_path": load_json_path,
+        "load_json_btn": load_json_btn,
+        "load_json_status": load_json_status,
+        "audio_directory": audio_directory,
+        "scan_btn": scan_btn,
+        "scan_status": scan_status,
+        "audio_files_table": audio_files_table,
+        "dataset_name": dataset_name,
+        "all_instrumental": all_instrumental,
+        "format_lyrics": format_lyrics,
+        "transcribe_lyrics": transcribe_lyrics,
+        "custom_tag": custom_tag,
+        "tag_position": tag_position,
+        "genre_ratio": genre_ratio,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_decomposition_contract_test.py b/acestep/ui/gradio/interfaces/training_decomposition_contract_test.py
new file mode 100644
index 00000000..6018778a
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_decomposition_contract_test.py
@@ -0,0 +1,154 @@
+"""AST contract tests for training interface decomposition."""
+
+from __future__ import annotations
+
+import ast
+import unittest
+from pathlib import Path
+
+try:
+    from .training_contract_ast_utils import (
+        call_name,
+        collect_return_dict_keys,
+        collect_training_section_keys_used_by_wiring,
+        load_module,
+    )
+except ImportError:
+    from training_contract_ast_utils import (  # type: ignore[no-redef]
+        call_name,
+        collect_return_dict_keys,
+        collect_training_section_keys_used_by_wiring,
+        load_module,
+    )
+
+
+class TrainingDecompositionContractTests(unittest.TestCase):
+    """Verify the training interface facade composes focused helper modules."""
+
+    def test_training_facade_imports_tab_helpers(self) -> None:
+        """``training.py`` should import dataset, LoRA, and LoKr helper modules."""
+
+        module = load_module("training.py")
+        imported_modules = []
+        for node in ast.walk(module):
+            if isinstance(node, ast.ImportFrom) and node.module:
+                imported_modules.append(node.module)
+
+        self.assertIn(
+            "acestep.ui.gradio.interfaces.training_dataset_builder_tab",
+            imported_modules,
+        )
+        self.assertIn("acestep.ui.gradio.interfaces.training_lora_tab", imported_modules)
+        self.assertIn("acestep.ui.gradio.interfaces.training_lokr_tab", imported_modules)
+
+    def test_training_facade_merges_helper_sections(self) -> None:
+        """``training.py`` should compose helper returns into one training-section map."""
+
+        module = load_module("training.py")
+        call_names: list[str] = []
+        update_calls = 0
+        for node in ast.walk(module):
+            if not isinstance(node, ast.Call):
+                continue
+            name = call_name(node.func)
+            if name:
+                call_names.append(name)
+            if isinstance(node.func, ast.Attribute) and node.func.attr == "update":
+                update_calls += 1
+
+        self.assertIn("create_dataset_builder_tab", call_names)
+        self.assertIn("create_training_lora_tab", call_names)
+        self.assertIn("create_training_lokr_tab", call_names)
+        self.assertGreaterEqual(update_calls, 4)
+
+    def test_dataset_builder_tab_delegates_to_section_builders(self) -> None:
+        """Dataset-builder facade should delegate to scan/label/preprocess builders."""
+
+        module = load_module("training_dataset_builder_tab.py")
+        call_names: list[str] = []
+        for node in ast.walk(module):
+            if isinstance(node, ast.Call):
+                name = call_name(node.func)
+                if name:
+                    call_names.append(name)
+
+        self.assertIn("build_dataset_scan_and_settings_controls", call_names)
+        self.assertIn("build_dataset_label_and_preview_controls", call_names)
+        self.assertIn("build_dataset_save_and_preprocess_controls", call_names)
+
+    def test_training_keys_cover_wiring_requirements(self) -> None:
+        """Returned training keys should cover all keys consumed by wiring modules."""
+
+        produced_keys: set[str] = set()
+        key_sources = [
+            ("training.py", "create_training_section"),
+            ("training_dataset_tab_scan_settings.py", "build_dataset_scan_and_settings_controls"),
+            (
+                "training_dataset_tab_label_preview.py",
+                "build_dataset_label_and_preview_controls",
+            ),
+            (
+                "training_dataset_tab_save_preprocess.py",
+                "build_dataset_save_and_preprocess_controls",
+            ),
+            ("training_lora_tab_dataset.py", "build_lora_dataset_and_adapter_controls"),
+            ("training_lora_tab_run_export.py", "build_lora_run_and_export_controls"),
+            ("training_lokr_tab_dataset.py", "build_lokr_dataset_and_adapter_controls"),
+            ("training_lokr_tab_run_export.py", "build_lokr_run_and_export_controls"),
+        ]
+        for module_name, function_name in key_sources:
+            produced_keys |= collect_return_dict_keys(module_name, function_name)
+
+        required_keys = collect_training_section_keys_used_by_wiring()
+        self.assertTrue(
+            required_keys.issubset(produced_keys),
+            f"Missing training_section keys: {sorted(required_keys - produced_keys)}",
+        )
+
+    def test_training_ui_markers_preserved(self) -> None:
+        """Key emoji UI markers should remain present after decomposition."""
+
+        interfaces_dir = Path(__file__).resolve().parent
+        expected_markers = {
+            "training.py": ["🎵 LoRA Training for ACE-Step"],
+            "training_dataset_tab_scan_settings.py": ["📂 Load Existing Dataset", "🔍 Scan New Directory"],
+            "training_dataset_tab_label_preview.py": ["🤖", "👀"],
+            "training_dataset_tab_save_preprocess.py": ["💾", "⚡"],
+            "training_lora_tab_dataset.py": ["📊", "⚙️"],
+            "training_lora_tab_run_export.py": ["🎛️", "📦"],
+            "training_lokr_tab_run_export.py": ["🎛️", "📦"],
+        }
+        for module_name, markers in expected_markers.items():
+            source = (interfaces_dir / module_name).read_text(encoding="utf-8")
+            for marker in markers:
+                self.assertIn(marker, source, f"Missing marker {marker!r} in {module_name}")
+
+    def test_lokr_helpers_use_i18n_translation_calls(self) -> None:
+        """LoKr modules should import and call ``t(...)`` for user-facing labels."""
+
+        for module_name in (
+            "training_lokr_tab.py",
+            "training_lokr_tab_dataset.py",
+            "training_lokr_tab_run_export.py",
+        ):
+            module = load_module(module_name)
+            imported_i18n = False
+            call_names: list[str] = []
+            for node in ast.walk(module):
+                if (
+                    isinstance(node, ast.ImportFrom)
+                    and node.module == "acestep.ui.gradio.i18n"
+                    and any(alias.name == "t" for alias in node.names)
+                ):
+                    imported_i18n = True
+                if isinstance(node, ast.Call):
+                    name = call_name(node.func)
+                    if name:
+                        call_names.append(name)
+
+            self.assertTrue(imported_i18n, f"{module_name} does not import t from i18n")
+            self.assertIn("t", call_names, f"{module_name} does not call t(...)")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/acestep/ui/gradio/interfaces/training_lokr_tab.py b/acestep/ui/gradio/interfaces/training_lokr_tab.py
new file mode 100644
index 00000000..7938d271
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lokr_tab.py
@@ -0,0 +1,25 @@
+"""LoKr training-tab facade for the Gradio training interface."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.help_content import create_help_button
+from acestep.ui.gradio.i18n import t
+from acestep.ui.gradio.interfaces.training_lokr_tab_dataset import (
+    build_lokr_dataset_and_adapter_controls,
+)
+from acestep.ui.gradio.interfaces.training_lokr_tab_run_export import (
+    build_lokr_run_and_export_controls,
+)
+
+
+def create_training_lokr_tab() -> dict[str, object]:
+    """Create the LoKr training tab and return component handles for wiring."""
+
+    with gr.Tab(t("training.tab_train_lokr")):
+        create_help_button("training_lokr")
+        tab_controls: dict[str, object] = {}
+        tab_controls.update(build_lokr_dataset_and_adapter_controls())
+        tab_controls.update(build_lokr_run_and_export_controls())
+    return tab_controls
diff --git a/acestep/ui/gradio/interfaces/training_lokr_tab_dataset.py b/acestep/ui/gradio/interfaces/training_lokr_tab_dataset.py
new file mode 100644
index 00000000..b07101ee
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lokr_tab_dataset.py
@@ -0,0 +1,98 @@
+"""LoKr tab dataset and adapter-setting controls."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_lokr_dataset_and_adapter_controls() -> dict[str, object]:
+    """Render LoKr dataset selector and adapter-parameter controls."""
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            gr.HTML(f"<h3>📊 {t('training.lokr_section_tensors')}</h3>")
+            gr.Markdown(t("training.lokr_tensor_selection_desc"))
+
+            lokr_training_tensor_dir = gr.Textbox(
+                label=t("training.preprocessed_tensors_dir"),
+                placeholder="./datasets/preprocessed_tensors",
+                value="./datasets/preprocessed_tensors",
+                info=t("training.preprocessed_tensors_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            lokr_load_dataset_btn = gr.Button(t("training.load_dataset_btn"), variant="secondary")
+
+            lokr_training_dataset_info = gr.Textbox(
+                label=t("training.dataset_info"),
+                interactive=False,
+                lines=3,
+            )
+
+        with gr.Column(scale=1):
+            gr.HTML(f"<h3>⚙️ {t('training.lokr_section_settings')}</h3>")
+
+            lokr_linear_dim = gr.Slider(
+                minimum=4,
+                maximum=256,
+                step=4,
+                value=64,
+                label=t("training.lokr_linear_dim"),
+                info=t("training.lokr_linear_dim_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_linear_alpha = gr.Slider(
+                minimum=4,
+                maximum=512,
+                step=4,
+                value=128,
+                label=t("training.lokr_linear_alpha"),
+                info=t("training.lokr_linear_alpha_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_factor = gr.Number(
+                label=t("training.lokr_factor"),
+                value=-1,
+                precision=0,
+                info=t("training.lokr_factor_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_decompose_both = gr.Checkbox(
+                label=t("training.lokr_decompose_both"),
+                value=False,
+                info=t("training.lokr_decompose_both_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_use_tucker = gr.Checkbox(
+                label=t("training.lokr_use_tucker"),
+                value=False,
+                info=t("training.lokr_use_tucker_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_use_scalar = gr.Checkbox(
+                label=t("training.lokr_use_scalar"),
+                value=False,
+                info=t("training.lokr_use_scalar_info"),
+                elem_classes=["has-info-container"],
+            )
+            lokr_weight_decompose = gr.Checkbox(
+                label=t("training.lokr_weight_decompose"),
+                value=True,
+                info=t("training.lokr_weight_decompose_info"),
+                elem_classes=["has-info-container"],
+            )
+
+    return {
+        "lokr_training_tensor_dir": lokr_training_tensor_dir,
+        "lokr_load_dataset_btn": lokr_load_dataset_btn,
+        "lokr_training_dataset_info": lokr_training_dataset_info,
+        "lokr_linear_dim": lokr_linear_dim,
+        "lokr_linear_alpha": lokr_linear_alpha,
+        "lokr_factor": lokr_factor,
+        "lokr_decompose_both": lokr_decompose_both,
+        "lokr_use_tucker": lokr_use_tucker,
+        "lokr_use_scalar": lokr_use_scalar,
+        "lokr_weight_decompose": lokr_weight_decompose,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_lokr_tab_run_export.py b/acestep/ui/gradio/interfaces/training_lokr_tab_run_export.py
new file mode 100644
index 00000000..40e4f1ea
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lokr_tab_run_export.py
@@ -0,0 +1,162 @@
+"""LoKr tab run and export controls."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_lokr_run_and_export_controls() -> dict[str, object]:
+    """Render LoKr training-run and export controls for the training tab."""
+
+    gr.HTML(f"<hr><h3>🎛️ {t('training.train_section_params')}</h3>")
+
+    with gr.Row():
+        lokr_learning_rate = gr.Number(
+            label=t("training.learning_rate"),
+            value=1e-3,
+            info=t("training.lokr_learning_rate_info"),
+            elem_classes=["has-info-container"],
+        )
+
+        lokr_train_epochs = gr.Slider(
+            minimum=1,
+            maximum=4000,
+            step=1,
+            value=500,
+            label=t("training.max_epochs"),
+        )
+
+        lokr_train_batch_size = gr.Slider(
+            minimum=1,
+            maximum=8,
+            step=1,
+            value=1,
+            label=t("training.batch_size"),
+        )
+
+        lokr_gradient_accumulation = gr.Slider(
+            minimum=1,
+            maximum=16,
+            step=1,
+            value=4,
+            label=t("training.gradient_accumulation"),
+        )
+
+    with gr.Row():
+        lokr_save_every_n_epochs = gr.Slider(
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=10,
+            label=t("training.save_every_n_epochs"),
+        )
+
+        lokr_training_shift = gr.Slider(
+            minimum=1.0,
+            maximum=5.0,
+            step=0.5,
+            value=3.0,
+            label=t("training.shift"),
+            info=t("training.shift_info"),
+            elem_classes=["has-info-container"],
+        )
+
+        lokr_training_seed = gr.Number(
+            label=t("training.seed"),
+            value=42,
+            precision=0,
+        )
+
+    with gr.Row():
+        lokr_output_dir = gr.Textbox(
+            label=t("training.output_dir"),
+            value="./lokr_output",
+            placeholder="./lokr_output",
+            info=t("training.lokr_output_dir_info"),
+            elem_classes=["has-info-container"],
+        )
+
+    gr.HTML("<hr>")
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            start_lokr_training_btn = gr.Button(
+                t("training.start_lokr_training_btn"),
+                variant="primary",
+                size="lg",
+            )
+        with gr.Column(scale=1):
+            stop_lokr_training_btn = gr.Button(
+                t("training.stop_training_btn"),
+                variant="stop",
+                size="lg",
+            )
+
+    lokr_training_progress = gr.Textbox(
+        label=t("training.training_progress"),
+        interactive=False,
+        lines=2,
+    )
+
+    with gr.Row():
+        lokr_training_log = gr.Textbox(
+            label=t("training.training_log"),
+            interactive=False,
+            lines=10,
+            max_lines=15,
+            scale=1,
+        )
+        lokr_training_loss_plot = gr.Plot(
+            label=t("training.lokr_training_loss_title"),
+            scale=1,
+        )
+
+    gr.HTML(f"<hr><h3>📦 {t('training.lokr_export_header')}</h3>")
+
+    with gr.Row():
+        lokr_export_path = gr.Textbox(
+            label=t("training.export_path"),
+            value="./lokr_output/final_lokr",
+            placeholder="./lokr_output/my_lokr",
+        )
+        export_lokr_btn = gr.Button(t("training.export_lokr_btn"), variant="secondary")
+
+    with gr.Row():
+        lokr_export_epoch = gr.Dropdown(
+            choices=[t("training.latest_auto")],
+            value=t("training.latest_auto"),
+            label=t("training.lokr_checkpoint_epoch"),
+            info=t("training.lokr_checkpoint_epoch_info"),
+            elem_classes=["has-info-container"],
+        )
+        refresh_lokr_export_epochs_btn = gr.Button(
+            t("training.refresh_epochs_btn"), variant="secondary"
+        )
+
+    lokr_export_status = gr.Textbox(
+        label=t("training.export_status"),
+        interactive=False,
+    )
+
+    return {
+        "lokr_learning_rate": lokr_learning_rate,
+        "lokr_train_epochs": lokr_train_epochs,
+        "lokr_train_batch_size": lokr_train_batch_size,
+        "lokr_gradient_accumulation": lokr_gradient_accumulation,
+        "lokr_save_every_n_epochs": lokr_save_every_n_epochs,
+        "lokr_training_shift": lokr_training_shift,
+        "lokr_training_seed": lokr_training_seed,
+        "lokr_output_dir": lokr_output_dir,
+        "start_lokr_training_btn": start_lokr_training_btn,
+        "stop_lokr_training_btn": stop_lokr_training_btn,
+        "lokr_training_progress": lokr_training_progress,
+        "lokr_training_log": lokr_training_log,
+        "lokr_training_loss_plot": lokr_training_loss_plot,
+        "lokr_export_path": lokr_export_path,
+        "lokr_export_epoch": lokr_export_epoch,
+        "refresh_lokr_export_epochs_btn": refresh_lokr_export_epochs_btn,
+        "export_lokr_btn": export_lokr_btn,
+        "lokr_export_status": lokr_export_status,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_lora_tab.py b/acestep/ui/gradio/interfaces/training_lora_tab.py
new file mode 100644
index 00000000..c8654244
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lora_tab.py
@@ -0,0 +1,36 @@
+"""LoRA training-tab facade for the Gradio training interface."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.help_content import create_help_button
+from acestep.ui.gradio.i18n import t
+from acestep.ui.gradio.interfaces.training_lora_tab_dataset import (
+    build_lora_dataset_and_adapter_controls,
+)
+from acestep.ui.gradio.interfaces.training_lora_tab_run_export import (
+    build_lora_run_and_export_controls,
+)
+
+
+def create_training_lora_tab(
+    *,
+    epoch_min: int,
+    epoch_step: int,
+    epoch_default: int,
+) -> dict[str, object]:
+    """Create the LoRA training tab and return component handles for wiring."""
+
+    with gr.Tab(t("training.tab_train_lora")):
+        create_help_button("training_train")
+        tab_controls: dict[str, object] = {}
+        tab_controls.update(build_lora_dataset_and_adapter_controls())
+        tab_controls.update(
+            build_lora_run_and_export_controls(
+                epoch_min=epoch_min,
+                epoch_step=epoch_step,
+                epoch_default=epoch_default,
+            )
+        )
+    return tab_controls
diff --git a/acestep/ui/gradio/interfaces/training_lora_tab_dataset.py b/acestep/ui/gradio/interfaces/training_lora_tab_dataset.py
new file mode 100644
index 00000000..dbbabd92
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lora_tab_dataset.py
@@ -0,0 +1,72 @@
+"""LoRA tab dataset and adapter-setting controls."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_lora_dataset_and_adapter_controls() -> dict[str, object]:
+    """Render LoRA dataset selector and adapter-parameter controls."""
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            gr.HTML(f"<h3>📊 {t('training.train_section_tensors')}</h3>")
+            gr.Markdown(t("training.train_tensor_selection_desc"))
+
+            training_tensor_dir = gr.Textbox(
+                label=t("training.preprocessed_tensors_dir"),
+                placeholder="./datasets/preprocessed_tensors",
+                value="./datasets/preprocessed_tensors",
+                info=t("training.preprocessed_tensors_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            load_dataset_btn = gr.Button(t("training.load_dataset_btn"), variant="secondary")
+
+            training_dataset_info = gr.Textbox(
+                label=t("training.dataset_info"),
+                interactive=False,
+                lines=3,
+            )
+
+        with gr.Column(scale=1):
+            gr.HTML(f"<h3>⚙️ {t('training.train_section_lora')}</h3>")
+
+            lora_rank = gr.Slider(
+                minimum=4,
+                maximum=256,
+                step=4,
+                value=64,
+                label=t("training.lora_rank"),
+                info=t("training.lora_rank_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            lora_alpha = gr.Slider(
+                minimum=4,
+                maximum=512,
+                step=4,
+                value=128,
+                label=t("training.lora_alpha"),
+                info=t("training.lora_alpha_info"),
+                elem_classes=["has-info-container"],
+            )
+
+            lora_dropout = gr.Slider(
+                minimum=0.0,
+                maximum=0.5,
+                step=0.05,
+                value=0.1,
+                label=t("training.lora_dropout"),
+            )
+
+    return {
+        "training_tensor_dir": training_tensor_dir,
+        "load_dataset_btn": load_dataset_btn,
+        "training_dataset_info": training_dataset_info,
+        "lora_rank": lora_rank,
+        "lora_alpha": lora_alpha,
+        "lora_dropout": lora_dropout,
+    }
diff --git a/acestep/ui/gradio/interfaces/training_lora_tab_run_export.py b/acestep/ui/gradio/interfaces/training_lora_tab_run_export.py
new file mode 100644
index 00000000..d75f7e05
--- /dev/null
+++ b/acestep/ui/gradio/interfaces/training_lora_tab_run_export.py
@@ -0,0 +1,166 @@
+"""LoRA tab run and export controls."""
+
+from __future__ import annotations
+
+import gradio as gr
+
+from acestep.ui.gradio.i18n import t
+
+
+def build_lora_run_and_export_controls(
+    *,
+    epoch_min: int,
+    epoch_step: int,
+    epoch_default: int,
+) -> dict[str, object]:
+    """Render LoRA training-run and export controls for the training tab."""
+
+    gr.HTML(f"<hr><h3>🎛️ {t('training.train_section_params')}</h3>")
+
+    with gr.Row():
+        learning_rate = gr.Number(
+            label=t("training.learning_rate"),
+            value=3e-4,
+            info=t("training.learning_rate_info"),
+            elem_classes=["has-info-container"],
+        )
+
+        train_epochs = gr.Slider(
+            minimum=epoch_min,
+            maximum=4000,
+            step=epoch_step,
+            value=epoch_default,
+            label=t("training.max_epochs"),
+        )
+
+        train_batch_size = gr.Slider(
+            minimum=1,
+            maximum=8,
+            step=1,
+            value=1,
+            label=t("training.batch_size"),
+            info=t("training.batch_size_info"),
+            elem_classes=["has-info-container"],
+        )
+
+        gradient_accumulation = gr.Slider(
+            minimum=1,
+            maximum=16,
+            step=1,
+            value=1,
+            label=t("training.gradient_accumulation"),
+            info=t("training.gradient_accumulation_info"),
+            elem_classes=["has-info-container"],
+        )
+
+    with gr.Row():
+        save_every_n_epochs = gr.Slider(
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=10,
+            label=t("training.save_every_n_epochs"),
+        )
+
+        training_shift = gr.Slider(
+            minimum=1.0,
+            maximum=5.0,
+            step=0.5,
+            value=3.0,
+            label=t("training.shift"),
+            info=t("training.shift_info"),
+            elem_classes=["has-info-container"],
+        )
+
+        training_seed = gr.Number(
+            label=t("training.seed"),
+            value=42,
+            precision=0,
+        )
+
+    with gr.Row():
+        lora_output_dir = gr.Textbox(
+            label=t("training.output_dir"),
+            value="./lora_output",
+            placeholder="./lora_output",
+            info=t("training.output_dir_info"),
+            elem_classes=["has-info-container"],
+        )
+
+    with gr.Row():
+        resume_checkpoint_dir = gr.Textbox(
+            label="Resume Checkpoint",
+            placeholder="./lora_output/checkpoints/epoch_200",
+            info="Directory of a saved LoRA checkpoint to resume from",
+            elem_classes=["has-info-container"],
+        )
+
+    gr.HTML("<hr>")
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            start_training_btn = gr.Button(
+                t("training.start_training_btn"),
+                variant="primary",
+                size="lg",
+            )
+        with gr.Column(scale=1):
+            stop_training_btn = gr.Button(
+                t("training.stop_training_btn"),
+                variant="stop",
+                size="lg",
+            )
+
+    training_progress = gr.Textbox(
+        label=t("training.training_progress"),
+        interactive=False,
+        lines=2,
+    )
+
+    with gr.Row():
+        training_log = gr.Textbox(
+            label=t("training.training_log"),
+            interactive=False,
+            lines=10,
+            max_lines=15,
+            scale=1,
+        )
+        training_loss_plot = gr.Plot(
+            label=t("training.training_loss_title"),
+            scale=1,
+        )
+
+    gr.HTML(f"<hr><h3>📦 {t('training.export_header')}</h3>")
+
+    with gr.Row():
+        export_path = gr.Textbox(
+            label=t("training.export_path"),
+            value="./lora_output/final_lora",
+            placeholder="./lora_output/my_lora",
+        )
+        export_lora_btn = gr.Button(t("training.export_lora_btn"), variant="secondary")
+
+    export_status = gr.Textbox(
+        label=t("training.export_status"),
+        interactive=False,
+    )
+
+    return {
+        "learning_rate": learning_rate,
+        "train_epochs": train_epochs,
+        "train_batch_size": train_batch_size,
+        "gradient_accumulation": gradient_accumulation,
+        "save_every_n_epochs": save_every_n_epochs,
+        "training_shift": training_shift,
+        "training_seed": training_seed,
+        "lora_output_dir": lora_output_dir,
+        "resume_checkpoint_dir": resume_checkpoint_dir,
+        "start_training_btn": start_training_btn,
+        "stop_training_btn": stop_training_btn,
+        "training_progress": training_progress,
+        "training_log": training_log,
+        "training_loss_plot": training_loss_plot,
+        "export_path": export_path,
+        "export_lora_btn": export_lora_btn,
+        "export_status": export_status,
+    }