added the max_matching_ngram_size to GenerationConfig (#29131)

* added the max_matching_ngram_size parameter into the GenerationConfig, for the PromptLookupCandidateGenerator

* switched back to keyword arguments

* added PromptLookupCandidateGenerator docstring for its parameters

* ruff reformat

* Update src/transformers/generation/configuration_utils.py
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

---------
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

src/transformers/generation/candidate_generator.py

@@ -252,10 +252,10 @@ class PromptLookupCandidateGenerator(CandidateGenerator):
     def __init__(
         self,
         num_output_tokens: int = 10,
-        max_matching_ngram_size: int = 2,
+        max_matching_ngram_size: int = None,
     ):
         self.num_output_tokens = num_output_tokens
-        self.max_matching_ngram_size = max_matching_ngram_size
+        self.max_matching_ngram_size = max_matching_ngram_size if max_matching_ngram_size else 2
 
         if self.max_matching_ngram_size <= 0 or self.num_output_tokens <= 0:
             raise ValueError("Invalid max_matching_ngram_size or num_output_tokens")

src/transformers/generation/configuration_utils.py

@@ -279,11 +279,18 @@ class GenerationConfig(PushToHubMixin):
             - `"heuristic_transient"`: Same as `"heuristic"` but `num_assistant_tokens` is reset to its initial value after each generation call.
             - `"constant"`: `num_assistant_tokens` stays unchanged during generation
 
+        prompt_lookup_num_tokens (`int`, *optional*, default to `None`):
+            The number of tokens to be output as candidate tokens.
+
+        max_matching_ngram_size (`int`, *optional*, default to `None`):
+            The maximum ngram size to be considered for matching in the prompt. Default to 2 if not provided.
+
         > Parameters specific to the caching mechanism:
 
         cache_implementation (`str`, *optional*, default to `None`):
             Cache class that should be used when generating.
 
+
         > Wild card
 
         generation_kwargs:
@@ -360,6 +367,7 @@ class GenerationConfig(PushToHubMixin):
 
         # Prompt lookup decoding
         self.prompt_lookup_num_tokens = kwargs.pop("prompt_lookup_num_tokens", None)
+        self.max_matching_ngram_size = kwargs.pop("max_matching_ngram_size", None)
 
         # Wild card
         self.generation_kwargs = kwargs.pop("generation_kwargs", {})

src/transformers/generation/utils.py

@@ -688,6 +688,7 @@ class GenerationMixin:
         if generation_config.prompt_lookup_num_tokens is not None:
             candidate_generator = PromptLookupCandidateGenerator(
                 num_output_tokens=generation_config.prompt_lookup_num_tokens,
+                max_matching_ngram_size=generation_config.max_matching_ngram_size,
             )
         else:
             candidate_generator = AssistedCandidateGenerator(