change text in Embeddings and Approximate KNN on GPU.ipynb

docs/Explore Algorithms/OpenAI/Quickstart - Custom Embeddings and Approximate KNN on GPU.ipynb

@@ -15,9 +15,10 @@
     }
    },
    "source": [
-    "# Embedding Text with local (per node) NVIDIA TensorRT accelerator and GPU based Aproximate Nearest Neighbor (ANN)\n",
+    "# Embedding with local (per node) NVIDIA TensorRT accelerator and GPU based Approximate Nearest Neighbor (ANN)\n",
     "\n",
-    "The demo extending existing [Azure OpenAI based demo](https://github.com/microsoft/SynapseML/blob/master/docs/Explore%20Algorithms/OpenAI/Quickstart%20-%20OpenAI%20Embedding%20and%20GPU%20based%20KNN.ipynb) when encoding is processed by OpenAI requests and KNN was using GPU based brute force search. This tutorial shows how to perform fast local embeddings using [multilingual E5 text embeddings](https://arxiv.org/abs/2402.05672) and fast aproximate Nearest Neighbor search using IVFFlat alcorithm. All tutorial stages accelerated by NVIDIA GPU using [NVIDIA TensorRT](https://developer.nvidia.com/tensorrt) and [Spark Rapids ML](https://github.com/NVIDIA/spark-rapids-ml). The tutorial folder contains two benchmark notebooks to demonstrate advantages of the presented GPU based approach compare to [previos CPU based demo](https://github.com/microsoft/SynapseML/blob/master/docs/Explore%20Algorithms/OpenAI/Quickstart%20-%20OpenAI%20Embedding.ipynb)\n",
+    "The demo extending existing [Azure OpenAI based demo](https://github.com/microsoft/SynapseML/blob/master/docs/Explore%20Algorithms/OpenAI/Quickstart%20-%20OpenAI%20Embedding%20and%20GPU%20based%20KNN.ipynb). Now encoding is processed by local embedders from Hugging Face and KNN is using GPU accelerated approximate method using IVFFlat algorithm. All tutorial stages accelerated by NVIDIA GPU using NVIDIA TensorRT and Spark Rapids ML.\n",
+    " All tutorial stages accelerated by NVIDIA GPU using [NVIDIA TensorRT](https://developer.nvidia.com/tensorrt) and [Spark Rapids ML](https://github.com/NVIDIA/spark-rapids-ml). The tutorial folder contains two benchmark notebooks to demonstrate advantages of the presented GPU based approach compare to [previous CPU based demo](https://github.com/microsoft/SynapseML/blob/master/docs/Explore%20Algorithms/OpenAI/Quickstart%20-%20OpenAI%20Embedding.ipynb)\n",
     "\n",
     "The key prerequisites for this quickstart include a working Azure OpenAI resource, and an Apache Spark cluster with SynapseML installed. We suggest creating a Synapse workspace, but currently the notebook was running on Databricks GPU based cluster using Standard_NC24ads_A100_v4 with 6 workers. Databricks Runtime was 13.3 LTS ML (includes Apache Spark 3.4.1, GPU, Scala 2.12) with related [init_script](https://github.com/microsoft/SynapseML/tree/master/tools/init_scripts) to install all required packages.\n"
    ]
@@ -466,7 +467,7 @@
     "\n",
     "The goal of this demo is to showcase two acceleration techniques: local (per node) embedding generation and approximate KNN. Compared to the original method, which relies on HTTP requests to the OpenAI model and CPU-based KNN. The new approach is significantly more scalable and provides substantial acceleration, especially for large input datasets.\n",
     "\n",
-    "This is the comparison dureation results on 10 T4 GPU nodes for both approaches:\n",
+    "This is the comparison duration results on 10 T4 GPU nodes for both approaches:\n",
     "\n",
     "![KNN Comparison](https://mmlspark.blob.core.windows.net/graphics/Documentation/knn_comparison.png)\n",
     "\n",