docs: clarify download stats; add usage with config.json

README.md

@@ -32,7 +32,7 @@ det_weights = torch.load(det_path, map_location="cpu")
 rec_weights = torch.load(rec_path, map_location="cpu")
 ```
 
-Direct links (also counted):
+Direct links (not counted by default):
 
 - https://huggingface.co/JoyCN/PaddleOCR-Pytorch/resolve/main/ptocr_v5_server_det.pth?download=true
 - https://huggingface.co/JoyCN/PaddleOCR-Pytorch/resolve/main/ptocr_v5_server_rec.pth?download=true
@@ -42,3 +42,20 @@ Direct links (also counted):
 - Please prefer the `huggingface_hub` API (`hf_hub_download`/`snapshot_download`) instead of `git lfs` cloning for downloads to be counted in Hub stats.
 - Stats update periodically; it may take some time for downloads to appear.
 
+## Download stats / 下载统计
+
+- Hub 统计依赖“查询文件”（query files），如 `config.json`；`.pth` 不是默认查询文件。
+- 若希望出现在统计中：先请求一次 `config.json`（GET/HEAD），再下载权重。
+
+Example:
+
+```python
+from huggingface_hub import hf_hub_download
+# Trigger stats by fetching the query file
+_ = hf_hub_download("JoyCN/PaddleOCR-Pytorch", filename="config.json")
+# Then download weights
+hf_hub_download("JoyCN/PaddleOCR-Pytorch", filename="ptocr_v5_server_det.pth")
+hf_hub_download("JoyCN/PaddleOCR-Pytorch", filename="ptocr_v5_server_rec.pth")
+```
+
+- 统计有延迟（常见数小时，最长 24–48 小时）。