Fix base_model target (Qwen/Qwen2.5-VL-7B-Instruct)

The `base_model` points to `Qwen/Qwen2.5-VL-7B`, which doesn't exist on the Hub — the actual upstream is [`Qwen/Qwen2.5-VL-7B-Instruct`](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) (matches the Qwen2.5-VL architecture this was built on). Because the current target is unresolvable, the repo currently appears as a *base* model with no parent link.

This corrects the target:

```yaml
base_model: Qwen/Qwen2.5-VL-7B-Instruct
```

It'll surface the model under Qwen/Qwen2.5-VL-7B-Instruct's **Model Tree** and improve discoverability. Metadata only.

Feel free to merge directly (or close).

README.md

@@ -1,6 +1,6 @@
 ---
 license: mit
-base_model: Qwen/Qwen2.5-VL-7B
+base_model: Qwen/Qwen2.5-VL-7B-Instruct
 new_version: numind/NuExtract3
 tags:
 - OCR