docs: fix custom model examples

docs/source/evaluating-a-custom-model.mdx

@@ -13,31 +13,46 @@ Create a Python file containing your custom model implementation. The model must
 Here's a basic example:
 
 ```python
+from transformers import AutoTokenizer
+
 from lighteval.models.abstract_model import LightevalModel
 from lighteval.models.model_output import ModelResponse
 from lighteval.tasks.requests import Doc, SamplingMethod
 from lighteval.utils.cache_management import SampleCache, cached
 
 class MyCustomModel(LightevalModel):
     def __init__(self, config):
-        super().__init__(config)
+        self.config = config
         # Initialize your model here...
+        self._tokenizer = AutoTokenizer.from_pretrained("gpt2")
 
         # Enable caching (recommended)
         self._cache = SampleCache(config)
 
+    @property
+    def tokenizer(self):
+        return self._tokenizer
+
+    @property
+    def add_special_tokens(self) -> bool:
+        return False
+
+    @property
+    def max_length(self) -> int:
+        return 2048
+
     @cached(SamplingMethod.GENERATIVE)
-    def greedy_until(self, docs: List[Doc]) -> List[ModelResponse]:
+    def greedy_until(self, docs: list[Doc]) -> list[ModelResponse]:
         # Implement generation logic
         pass
 
     @cached(SamplingMethod.LOGPROBS)
-    def loglikelihood(self, docs: List[Doc]) -> List[ModelResponse]:
+    def loglikelihood(self, docs: list[Doc]) -> list[ModelResponse]:
         # Implement loglikelihood computation
         pass
 
     @cached(SamplingMethod.PERPLEXITY)
-    def loglikelihood_rolling(self, docs: List[Doc]) -> List[ModelResponse]:
+    def loglikelihood_rolling(self, docs: list[Doc]) -> list[ModelResponse]:
         # Implement rolling loglikelihood computation
         pass
 ```
@@ -59,7 +74,7 @@ You can evaluate your custom model using either the command-line interface or th
 lighteval custom \
     "google-translate" \
     "examples/custom_models/google_translate_model.py" \
-    "wmt20:fr-de \
+    "lighteval|wmt20:fr-de|0" \
     --max-samples 10
 ```
 
@@ -94,7 +109,7 @@ model_config = CustomModelConfig(
 
 # Create and run the pipeline
 pipeline = Pipeline(
-    tasks=truthfulqa:mc,
+    tasks="truthfulqa:mc",
     pipeline_parameters=pipeline_params,
     evaluation_tracker=evaluation_tracker,
     model_config=model_config
@@ -174,17 +189,18 @@ from lighteval.utils.cache_management import SampleCache, cached
 ### Step 2: Initialize Cache in Constructor
 ```python
 def __init__(self, config):
-    super().__init__(config)
+    self.config = config
     # Your initialization code...
     self._cache = SampleCache(config)
 ```
 
-3. Add cache decorators to your prediction methods:
-   ```python
-   @cached(SamplingMethod.GENERATIVE)
-   def greedy_until(self, docs: List[Doc]) -> List[ModelResponse]:
-       # Your implementation...
-   ```
+### Step 3: Add Cache Decorators
+```python
+@cached(SamplingMethod.GENERATIVE)
+def greedy_until(self, docs: list[Doc]) -> list[ModelResponse]:
+    # Your implementation...
+    pass
+```
 
 For detailed information about the caching system, see the [Caching Documentation](caching).
 

src/lighteval/models/custom/custom_model.py

@@ -33,7 +33,7 @@ class CustomModelConfig(ModelConfig):
     This class will be automatically detected and instantiated when loading the model.
 
     Args:
-        model (str):
+        model_name (str):
             An identifier for the model. This can be used to track which model was evaluated
             in the results and logs.
 
@@ -46,24 +46,42 @@ class CustomModelConfig(ModelConfig):
         ```python
         # Define config
         config = CustomModelConfig(
-            model="my-custom-model",
+            model_name="my-custom-model",
             model_definition_file_path="path/to/my_model.py"
         )
 
         # Example custom model file (my_model.py):
         from lighteval.models.abstract_model import LightevalModel
+        from lighteval.models.model_output import ModelResponse
+        from lighteval.tasks.requests import Doc
 
         class MyCustomModel(LightevalModel):
-            def __init__(self, config, env_config):
-                super().__init__(config, env_config)
+            def __init__(self, config):
+                self.config = config
                 # Custom initialization...
 
+            @property
+            def tokenizer(self):
+                # Return the tokenizer used by your model
+                ...
+
+            @property
+            def add_special_tokens(self) -> bool:
+                return False
+
+            @property
+            def max_length(self) -> int:
+                return 2048
+
             def greedy_until(self, docs: list[Doc]) -> list[ModelResponse]:
                 # Custom generation logic...
                 pass
 
             def loglikelihood(self, docs: list[Doc]) -> list[ModelResponse]:
                 pass
+
+            def loglikelihood_rolling(self, docs: list[Doc]) -> list[ModelResponse]:
+                pass
         ```
 
     An example of a custom model can be found in `examples/custom_models/google_translate_model.py`.