document third party exllamav2 with logits processor

docs/reference/models/exllamav2.md

@@ -1,7 +1,15 @@
 # ExllamaV2
 
+The `outlines.models.exllamav2` model requires a Logits Processor component for compatibility with Outlines structured generation. While ExLlamaV2 doesn't natively support this feature, a third-party fork provides the necessary functionality. You can install it with the following command:
+
+```bash
+pip install git+https://github.com/lapp0/exllamav2@sampler-logits-processor
+```
+
+Install other requirements:
+
 ```bash
-pip install exllamav2 transformers torch
+pip install transformers torch
 ```
 
 *Coming soon*

outlines/models/exllamav2.py

@@ -300,7 +300,9 @@ def exl2(
 
     except ImportError:
         raise ImportError(
-            "The `exllamav2`, `transformers` and `torch` libraries needs to be installed in order to use `exllamav2` models."
+            "The `exllamav2`, `transformers` and `torch` libraries needs to be installed in order to use `exllamav2` models. "
+            "Please run `pip install transformers torch git+https://github.com/lapp0/exllamav2@sampler-logits-processor` "
+            "Documentation: https://dottxt-ai.github.io/outlines/reference/models/exllamav2/"
         )
     config = ExLlamaV2Config(model_path)
     if max_chunk_size is not None: