🌀 Fix GRPO default completion length doc (#2662)

huggingface · Jan 26, 2025 · 55a329e · 55a329e
1 parent 4720656
commit 55a329e
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/trl/trainer/grpo_config.py b/trl/trainer/grpo_config.py
@@ -48,7 +48,7 @@ class GRPOConfig(TrainingArguments):
             Number of generations per prompt to sample.
         temperature (`float`, *optional*, defaults to `0.9`):
             Temperature for sampling. The higher the temperature, the more random the completions.
-        max_completion_length (`int` or `None`, *optional*, defaults to `None`):
+        max_completion_length (`int` or `None`, *optional*, defaults to `256`):
             Maximum length of the generated completion.
 
         > Parameters that control the training