Rename models_manager_v8.2.py to models_manager_v8.3.py

models_manager_v8.2.py → models_manager_v8.3.py

@@ -631,18 +631,43 @@ class ModelsManager(tk.Tk):
 
         ttk.Button(win, text="自动生成 dataset", command=auto_dataset).pack(anchor="w", padx=10, pady=4)
         
+        # # 🌟 新增：动态渲染自定义字段
+        # custom_vars = {}
+        # locked_keys = ["title", "trigger", "dataset", "showcase"]
+        # custom_keys = [k for k in meta.keys() if k not in locked_keys]
+        
+        # if custom_keys:
+        #     ext_frame = ttk.LabelFrame(win, text="自定义扩展字段")
+        #     ext_frame.pack(fill="x", padx=10, pady=5)
+        #     for k in custom_keys:
+        #         row = ttk.Frame(ext_frame)
+        #         row.pack(fill="x", pady=2)
+        #         ttk.Label(row, text=f"{k}:", width=12).pack(side="left")
+        #         var = tk.StringVar(value=str(meta.get(k, "")))
+        #         custom_vars[k] = var
+        #         ttk.Entry(row, textvariable=var).pack(side="left", fill="x", expand=True, padx=5)
         # ========================================================
-        # 🌟 新增：渲染动态扩展字段
+        # 🌟 动态渲染自定义字段（兼容空数据）
         # ========================================================
-        # 🌟 新增：动态渲染自定义字段
         custom_vars = {}
         locked_keys = ["title", "trigger", "dataset", "showcase"]
-        custom_keys = [k for k in meta.keys() if k not in locked_keys]
         
-        if custom_keys:
+        # 1. 当前数据自身的 key
+        current_custom_keys = set([k for k in meta.keys() if k not in locked_keys])
+
+        # 2. 全局所有数据里出现过的 key
+        global_custom_keys = set()
+        for m in self.items.values():
+            if isinstance(m, dict):
+                global_custom_keys.update([k for k in m.keys() if k not in locked_keys])
+
+        # 3. 合并去重
+        all_custom_keys = list(current_custom_keys) + [k for k in global_custom_keys if k not in current_custom_keys]
+
+        if all_custom_keys:
             ext_frame = ttk.LabelFrame(win, text="自定义扩展字段")
             ext_frame.pack(fill="x", padx=10, pady=5)
-            for k in custom_keys:
+            for k in all_custom_keys:
                 row = ttk.Frame(ext_frame)
                 row.pack(fill="x", pady=2)
                 ttk.Label(row, text=f"{k}:", width=12).pack(side="left")
@@ -1297,7 +1322,7 @@ class ModelsManager(tk.Tk):
                 self._convert_png_to_jpg(local_dir)
                 
                 # 4. 打包 Parquet 并更新 YAML
-                temp_readme_path, valid_folders = self._generate_parquet_and_update_yaml(api, local_dir, region_prefix)
+                temp_readme_path, valid_folders = self._generate_parquet_and_update(api, local_dir, region_prefix)
                 
                 # 5. 上传
                 self._upload_files_and_readme(api, local_dir, region_prefix, temp_readme_path, valid_folders)
@@ -1407,40 +1432,12 @@ class ModelsManager(tk.Tk):
                 return
         self.log_dataset(f"  🎉 转换完成！共转换 {converted_count} 张图片。")
 
-    def _generate_parquet_and_update_yaml(self, api, current_dir, region_prefix):
-        import glob, tempfile, pandas as pd
+    def _generate_parquet_and_update(self, api, current_dir, region_prefix):
+        import glob, pandas as pd
         from datasets import Dataset, Features, Image as HFImage, Value
-        from huggingface_hub import DatasetCard
-        from PIL import Image as PILImage
-
-        self.log_dataset("📦 [任务 4/5] 正在本地打包 Parquet 并更新 YAML 配置...")
-        local_folder_name = os.path.basename(current_dir).replace(" ", "_").replace("-", "_")
-        
-        self.log_dataset(f"  ☁️ 正在读取云端 {DATASET_REPO_ID} 的现有配置...")
-        try:
-            # 注意这里用 DATASET_REPO_ID 替换了原脚本的 REPO_ID
-            remote_readme_path = api.hf_hub_download(repo_id=DATASET_REPO_ID, filename="README.md", repo_type="dataset")
-            with open(remote_readme_path, "r", encoding="utf-8") as f:
-                content = f.read()
-            card = DatasetCard(content)
-            self.log_dataset("  ✅ 成功读取云端 README。")
-        except Exception as e:
-            self.log_dataset(f"  ⚠️ 找不到云端 README，将新建: {e}")
-            card = DatasetCard("")
 
-        yaml_data = card.data if card.data else {}
-        if "configs" not in yaml_data: yaml_data["configs"] = []
+        self.log_dataset("📦 [任务 4/5] 正在本地打包 Parquet (已启用极简聚合模式)...")
         
-        yaml_data["dataset_info"] = {
-            "features": [
-                {"name": "image", "dtype": "image"},
-                {"name": "file_name", "dtype": "string"},
-                {"name": "text", "dtype": "string"}
-            ]
-        }
-
-        existing_config_names = [c.get("config_name") for c in yaml_data["configs"]]
-        new_configs_added = False
         valid_folders = []
 
         for sub_folder_name in os.listdir(current_dir):
@@ -1489,32 +1486,6 @@ class ModelsManager(tk.Tk):
                     valid_folders.append(sub_folder_name)
                     self.log_dataset(f"    ✅ 生成 train.parquet ({len(data_rows)} 条数据)")
 
-                    config_name = f"{local_folder_name}_{sub_folder_name}".lower()
-                    path_str = f"{region_prefix}/{local_folder_name}/{sub_folder_name}/train.parquet"
-                    
-                    if config_name not in existing_config_names:
-                        yaml_data["configs"].append({
-                            "config_name": config_name,
-                            "data_files": [{"split": "train", "path": path_str}]
-                        })
-                        new_configs_added = True
-                        self.log_dataset(f"    ➕ 已添加配置: {config_name}")
-
-        def get_priority(config):
-            data_files = config.get("data_files", [])
-            path = data_files[0].get("path", "") if data_files else ""
-            priority_dict = {"CN": 1, "KR": 2, "JP": 3, "Fictional": 4, "WE": 5, "Heels": 6,}
-            path_prefix = path.split('/')[0] if path else ""
-            return priority_dict.get(path_prefix, 99)
-
-        yaml_data["configs"] = sorted(yaml_data["configs"], key=lambda c: (get_priority(c), c.get("config_name", "")))
-        self.log_dataset("  🔄 已根据 [CN > KR > JP > Fictional > WE > Heels] 的顺序重排 YAML。")
-
-        if new_configs_added:
-            card.data = yaml_data
-            with tempfile.NamedTemporaryFile(mode='w', encoding='utf-8', suffix='.md', delete=False) as temp_readme:
-                temp_readme.write(str(card))
-                return temp_readme.name, valid_folders
         return None, valid_folders
 
     def _upload_files_and_readme(self, api, current_dir, region_prefix, temp_readme_path, valid_folders):
@@ -1538,22 +1509,6 @@ class ModelsManager(tk.Tk):
                 self.log_dataset(f"  🎉 {sub_folder_name} 上传完成！")
             except Exception as e:
                 self.log_dataset(f"  ❌ {sub_folder_name} 上传失败: {e}")
-
-        if temp_readme_path:
-            self.log_dataset("\n  📝 正在推送最新的 README.md...")
-            try:
-                api.upload_file(
-                    path_or_fileobj=temp_readme_path,
-                    path_in_repo="README.md",
-                    repo_id=DATASET_REPO_ID,
-                    repo_type="dataset",
-                    commit_message="Auto-update README YAML configs"
-                )
-                self.log_dataset("  🎉 README.md 已成功更新到云端！")
-            except Exception as e:
-                self.log_dataset(f"  ❌ README.md 上传失败: {e}")
-            finally:
-                os.remove(temp_readme_path)
     # ==========================================================
     # ==================== 区域 4: Showcase 管理 ================
     # ==========================================================