update doc

Signed-off-by: Yuki Huang <yukih@nvidia.com>

Author: yuki-666

docs/design-docs/fsdp2-parallel-plan.md

@@ -4,7 +4,7 @@ This guide outlines the parallelization strategy for FSDP2 training in NeMo-RL.
 
 ## Fallback Priority
 
-Three parallelization approaches are supported, which is falling back in this priority.
+Three parallelization approaches are supported, with the following fallback priority.
 
 **Custom Parallel Plan**
 
@@ -14,13 +14,13 @@ For implementation details and usage guidelines, please refer to [Custom Paralle
 
 **Optimized Parallel Plan**
 
-Optimized parallel plans are available for specific model architectures, potentially offering superior performance compared to the Hugging Face tensor parallel implementation.
+Optimized parallel plans are available for specific model architectures and may offer superior performance compared to the Hugging Face tensor parallel implementation.
 
-This approach is employed when no custom parallel plan is specified and the model class supports optimized parallelization.
+This approach is used when no custom parallel plan is specified and the model class supports optimized parallelization.
 
 **Hugging Face Tensor Parallel Plan**
 
-Hugging Face provides tensor parallel of most models through `._tp_plan`.
+Hugging Face provides tensor parallelism for most models through `._tp_plan`.
 
 It serves as the default when neither custom nor optimized parallel plans are available.