Update 5.mdx

chapters/en/chapter12/5.mdx

@@ -74,7 +74,7 @@ tokenizer = AutoTokenizer.from_pretrained(model_id)
 
 Now, let's load the LoRA configuration. We'll take advantage of LoRA to reduce the number of trainable parameters, and in turn the memory footprint we need to fine-tune the model.
 
-If you're not familiar with LoRA, you can read more about it in [Chapter 11](https://huggingface.co/learn/course/en/chapter11/3).
+If you're not familiar with LoRA, you can read more about it in [Chapter 11](https://huggingface.co/learn/llm-course/chapter11/4).
 
 ```python
 # Load LoRA
@@ -109,7 +109,7 @@ def reward_len(completions, **kwargs):
 
 Now, let's define the training arguments. We'll use the `GRPOConfig` class to define the training arguments in a typical `transformers` style.
 
-If this is the first time you're defining training arguments, you can check the [TrainingArguments](https://huggingface.co/docs/transformers/en/main_classes/trainer#trainingarguments) class for more information, or [Chapter 2](https://huggingface.co/learn/course/en/chapter2/1) for a detailed introduction.
+If this is the first time you're defining training arguments, you can check the [TrainingArguments](https://huggingface.co/docs/transformers/en/main_classes/trainer#trainingarguments) class for more information, or [Chapter 2](https://huggingface.co/learn/llm-course/en/chapter2/1) for a detailed introduction.
 
 ```python
 # Training arguments