feat: complete i18n with separate translation file

i18n.py

@@ -0,0 +1,174 @@
+"""Internationalization strings for VoxCPM Training WebUI.
+
+All translatable UI strings are defined here. Widget creation in
+lora_ft_webui.py uses t("key") for initial display; change_language()
+uses STRINGS[lang]["key"] to switch all elements.
+
+To add a new language: add a new key to STRINGS with all the same keys.
+"""
+
+import os
+
+DEFAULT_LANG = os.environ.get("VOXCPM_LANG", "zh")
+
+STRINGS = {
+    "zh": {
+        "title": "VoxCPM LoRA WebUI",
+        "tab_train": "训练 (Training)",
+        "tab_infer": "推理 (Inference)",
+        "pretrained_path": "预训练模型路径",
+        "train_manifest": "训练数据清单 (jsonl)",
+        "val_manifest": "验证数据清单 (可选)",
+        "lr": "学习率 (Learning Rate)",
+        "max_iters": "最大迭代次数",
+        "batch_size": "批次大小 (Batch Size)",
+        "lora_rank": "LoRA Rank",
+        "lora_alpha": "LoRA Alpha",
+        "save_interval": "保存间隔 (Steps)",
+        "start_train": "开始训练",
+        "stop_train": "停止训练",
+        "train_logs": "训练日志",
+        "text_to_synth": "合成文本",
+        "voice_cloning": "### 声音克隆 (可选)",
+        "ref_audio": "参考音频",
+        "ref_text": "参考文本 (可选)",
+        "select_lora": "选择 LoRA 模型",
+        "cfg_scale": "CFG Scale (引导系数)",
+        "infer_steps": "推理步数",
+        "seed": "随机种子 (Seed)",
+        "gen_audio": "生成音频",
+        "gen_output": "生成结果",
+        "status": "状态",
+        "lang_select": "Language / 语言",
+        "refresh": "刷新",
+        "output_name": "输出目录名称 (可选，若存在则继续训练)",
+        "grad_accum_steps": "梯度累积 (grad_accum_steps)",
+        "num_workers": "数据加载线程 (num_workers)",
+        "log_interval": "日志间隔 (log_interval)",
+        "valid_interval": "验证间隔 (valid_interval)",
+        "weight_decay": "权重衰减 (weight_decay)",
+        "warmup_steps": "预热步数 (warmup_steps)",
+        "max_steps": "最大步数 (max_steps)",
+        "sample_rate": "采样率 (sample_rate)",
+        "max_grad_norm": "梯度裁剪 (max_grad_norm, 0=关闭)",
+        "enable_lm": "启用 LoRA LM (enable_lm)",
+        "enable_dit": "启用 LoRA DIT (enable_dit)",
+        "enable_proj": "启用投影 (enable_proj)",
+        "dropout": "LoRA 丢弃率 (Dropout)",
+        "tensorboard_path": "Tensorboard 路径 (可选)",
+        "hf_model_id": "HuggingFace Model ID (e.g., openbmb/VoxCPM2)",
+        "distribute": "分发模式 (distribute)",
+        "header_title": "### 强大的语音合成和 LoRA 微调工具",
+        "header_subtitle": "支持语音克隆、LoRA 模型训练和推理的完整解决方案",
+        "train_setup_title": "### 🎯 模型训练设置",
+        "train_setup_desc": "配置你的 LoRA 微调训练参数",
+        "base_config_header": "#### 📁 基础配置",
+        "train_params_header": "#### ⚙️ 训练参数",
+        "advanced_accordion": "🔧 高级选项 (Advanced)",
+        "distribution_header": "#### 分发选项 (Distribution)",
+        "train_logs_header": "#### 📊 训练日志",
+        "infer_title": "### 🎤 语音合成",
+        "infer_desc": "使用训练好的 LoRA 模型生成语音，支持 LoRA 微调和声音克隆",
+        "input_config_header": "#### 📝 输入配置",
+        "voice_clone_header": "**🎭 声音克隆（可选）**",
+        "model_select_header": "#### 🤖 模型选择",
+        "gen_params_header": "#### ⚙️ 生成参数",
+        "gen_output_header": "#### 🎧 生成结果",
+        "status_header": "#### 📋 状态信息",
+        "placeholder_synth": "输入要合成的文本内容...",
+        "placeholder_ref_text": "如不填写，将自动识别参考音频内容",
+        "placeholder_status": "等待生成...",
+        "info_cfg": "引导系数，值越大越贴近提示",
+        "info_steps": "生成质量与步数成正比，但耗时更长",
+        "info_seed": "-1 为随机，固定值可复现结果",
+        "info_lora_select": "选择训练好的 LoRA 模型，或选择 None 使用基础模型",
+        "log_ref_audio_no_text": "参考音频已提供但缺少文本，自动识别中...",
+        "log_auto_recognized": "自动识别文本: {text}",
+        "error_asr_failed": "错误：无法识别参考音频内容，请手动填写参考文本",
+        "error_asr_exception": "错误：自动识别参考音频失败 - {error}",
+        "log_refresh_lora": "刷新 LoRA 列表: 找到 {count} 个检查点",
+    },
+    "en": {
+        "title": "VoxCPM LoRA WebUI",
+        "tab_train": "Training",
+        "tab_infer": "Inference",
+        "pretrained_path": "Pretrained Model Path",
+        "train_manifest": "Train Manifest (jsonl)",
+        "val_manifest": "Validation Manifest (Optional)",
+        "lr": "Learning Rate",
+        "max_iters": "Max Iterations",
+        "batch_size": "Batch Size",
+        "lora_rank": "LoRA Rank",
+        "lora_alpha": "LoRA Alpha",
+        "save_interval": "Save Interval",
+        "start_train": "▶️ Start Training",
+        "stop_train": "⏹️ Stop Training",
+        "train_logs": "Training Logs",
+        "text_to_synth": "Text to Synthesize",
+        "voice_cloning": "### Voice Cloning (Optional)",
+        "ref_audio": "Reference Audio",
+        "ref_text": "Reference Text (Optional)",
+        "select_lora": "Select LoRA Checkpoint",
+        "cfg_scale": "CFG Scale",
+        "infer_steps": "Inference Steps",
+        "seed": "Seed",
+        "gen_audio": "🎵 Generate Audio",
+        "gen_output": "Generated Audio",
+        "status": "Status",
+        "lang_select": "Language / 语言",
+        "refresh": "🔄 Refresh Models",
+        "output_name": "Output Name (Optional, resume if exists)",
+        "header_title": "### Powerful voice synthesis and LoRA fine-tuning",
+        "header_subtitle": "Complete solution for voice cloning, LoRA model training and inference",
+        "train_setup_title": "### 🎯 Model Training Setup",
+        "train_setup_desc": "Configure your LoRA fine-tuning parameters",
+        "base_config_header": "#### 📁 Base Configuration",
+        "train_params_header": "#### ⚙️ Training Parameters",
+        "advanced_accordion": "🔧 Advanced Options",
+        "distribution_header": "#### Distribution Options",
+        "train_logs_header": "#### 📊 Training Logs",
+        "infer_title": "### 🎤 Voice Synthesis",
+        "infer_desc": "Generate speech with trained LoRA models — supports fine-tuning and voice cloning",
+        "input_config_header": "#### 📝 Input Configuration",
+        "voice_clone_header": "**🎭 Voice Cloning (Optional)**",
+        "model_select_header": "#### 🤖 Model Selection",
+        "gen_params_header": "#### ⚙️ Generation Parameters",
+        "gen_output_header": "#### 🎧 Generated Output",
+        "status_header": "#### 📋 Status",
+        "placeholder_synth": "Enter text to synthesize...",
+        "placeholder_ref_text": "If empty, auto-recognizes from reference audio",
+        "placeholder_status": "Waiting for generation...",
+        "info_cfg": "Guidance scale — higher values follow the prompt more closely",
+        "info_steps": "Higher = better quality but slower",
+        "info_seed": "-1 for random, fixed value for reproducible results",
+        "info_lora_select": "Select a trained LoRA model, or None for base model",
+        "grad_accum_steps": "Grad Accum Steps",
+        "num_workers": "Num Workers",
+        "log_interval": "Log Interval",
+        "valid_interval": "Valid Interval",
+        "weight_decay": "Weight Decay",
+        "warmup_steps": "Warmup Steps",
+        "max_steps": "Max Steps",
+        "sample_rate": "Sample Rate",
+        "max_grad_norm": "Max Grad Norm (0=disabled)",
+        "enable_lm": "Enable LoRA LM",
+        "enable_dit": "Enable LoRA DIT",
+        "enable_proj": "Enable Projection",
+        "dropout": "LoRA Dropout",
+        "tensorboard_path": "Tensorboard Path (Optional)",
+        "hf_model_id": "HuggingFace Model ID (e.g., openbmb/VoxCPM2)",
+        "distribute": "Distribute Mode",
+        "log_ref_audio_no_text": "Reference audio provided without text, auto-recognizing...",
+        "log_auto_recognized": "Auto-recognized text: {text}",
+        "error_asr_failed": "Error: Could not recognize reference audio, please enter reference text manually",
+        "error_asr_exception": "Error: Reference audio recognition failed - {error}",
+        "log_refresh_lora": "Refresh LoRA list: found {count} checkpoints",
+    },
+}
+
+
+def t(key: str, lang: str = None) -> str:
+    """Get translated string. Uses DEFAULT_LANG if lang not specified."""
+    if lang is None:
+        lang = DEFAULT_LANG
+    return STRINGS.get(lang, STRINGS["en"]).get(key, STRINGS["en"].get(key, key))