Update README.md

Author: g-husam

docs/README.md

@@ -23,7 +23,7 @@ See the [overview](overview.md) for more detail.
 
 ### Performance
 
-We performed some tests on a [Vertex AI Training Cluster](https://docs.cloud.google.com/vertex-ai/docs/training/training-clusters/overview) with 4 [A3-Mega](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines#a3-mega-vms) nodes for Gemma 27B and Llama 70B pre-training over just 300 steps.
+We performed some tests on a [Vertex AI Training Cluster](https://docs.cloud.google.com/vertex-ai/docs/training/training-clusters/overview) with 4 [A3-Mega](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines#a3-mega-vms) nodes for Gemma 27B and Llama 70B pre-training over just 300 steps and observed the improvements listed below.
 These tests were conducted using ML Flashpoint _alongside_ NeMo's recommended checkpointing (as you would in production), where NeMo's default checkpointing used a 7-10 TB [Filestore](https://cloud.google.com/filestore) instance.
 
 Observations when comparing the hybrid of ML Flashpoint (every 5 steps) and NeMo checkpointing (every 50 steps) to just NeMo's regular checkpointing (every 10 steps):