diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
new file mode 100644
index 0000000000..896912f3d1
--- /dev/null
+++ b/CONTRIBUTING.md
@@ -0,0 +1,33 @@
+---
+nav_exclude: true
+---
+## Developing
+
+Once you've installed dependencies with `npm install` (or  `yarn`), start a development server with hot-reload enabled:
+
+```bash
+npm run dev
+
+# or start the server and open the app in a new browser tab
+npm run dev -- --open
+```
+All working pages are in `src/routes/[page url]/+page.svelte`, which is where you can make your edits.
+
+### Technologies & relevant docs
+Please use the docs pages below to aid in your development process. As a general target, we should be using zero CSS, as daisyUI (framework with components) and tailwindcss (css classes) should be able to handle all of our styling needs.
+- [Svelte](https://svelte.dev/)
+- daisyUI [docs](https://daisyui.com/)
+- tailwindcss [docs](https://tailwindcss.com/docs)
+
+
+## Building
+
+To create a production version of your app:
+
+```bash
+npm run build
+```
+
+You can preview the production build with `npm run preview`.
+
+> To deploy your app, you may need to install an [adapter](https://kit.svelte.dev/docs/adapters) for your target environment.
diff --git a/README.md b/README.md
deleted file mode 100644
index e8bfc25c4e..0000000000
--- a/README.md
+++ /dev/null
@@ -1,34 +0,0 @@
-## Creating a project
-
-If you're seeing this, you've probably already done this step. Congrats!
-
-```bash
-# create a new project in the current directory
-npm create svelte@latest
-
-# create a new project in my-app
-npm create svelte@latest my-app
-```
-
-## Developing
-
-Once you've created a project and installed dependencies with `npm install` (or `pnpm install` or `yarn`), start a development server:
-
-```bash
-npm run dev
-
-# or start the server and open the app in a new browser tab
-npm run dev -- --open
-```
-
-## Building
-
-To create a production version of your app:
-
-```bash
-npm run build
-```
-
-You can preview the production build with `npm run preview`.
-
-> To deploy your app, you may need to install an [adapter](https://kit.svelte.dev/docs/adapters) for your target environment.
diff --git a/_config.yml b/_config.yml
index 0620f40ac8..3079a71fb0 100644
--- a/_config.yml
+++ b/_config.yml
@@ -35,3 +35,8 @@ gh_edit_repository: 'https://github.com/microsoft/onnxruntime' # the github URL
 gh_edit_branch: 'gh-pages' # the branch that your docs is served from
 # gh_edit_source: docs # the source that your files originate from
 gh_edit_view_mode: 'tree' # "tree" or "edit" if you want the user to jump into the editor immediately
+nav_external_links:
+  - title: ONNX Runtime Docs on GitHub
+    url: https://github.com/microsoft/onnxruntime/tree/gh-pages
+    hide_icon: true # set to true to hide the external link icon - defaults to false
+    opens_in_new_tab: true # set to true to open this link in a new tab - defaults to false
\ No newline at end of file
diff --git a/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png b/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png
index 4646c06045..c1da7555ea 100644
Binary files a/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png and b/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png b/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png
index eecce4751c..e777913981 100644
Binary files a/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png and b/src/images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png b/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png
index b8cdf011df..9b46047d21 100644
Binary files a/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png and b/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency 1.png b/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency 1.png
deleted file mode 100644
index 814a8e6d81..0000000000
Binary files a/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency 1.png and /dev/null differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency.png b/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency.png
new file mode 100644
index 0000000000..b950cff87f
Binary files /dev/null and b/src/images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency.png differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png b/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png
index b5ea1e1c5f..cfcc3016ad 100644
Binary files a/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png and b/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png differ
diff --git a/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png b/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png
index 0bc90a464b..9ba1c57c5b 100644
Binary files a/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png and b/src/images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png differ
diff --git a/src/routes/blogs/accelerating-llama-2/+page.svelte b/src/routes/blogs/accelerating-llama-2/+page.svelte
index 0f5add02a8..e8bec9b4fc 100644
--- a/src/routes/blogs/accelerating-llama-2/+page.svelte
+++ b/src/routes/blogs/accelerating-llama-2/+page.svelte
@@ -1,13 +1,13 @@
 <script>
 	import Header from '../../components/header.svelte';
 	import Footer from '../../components/footer.svelte';
-    import figure1 from '../../../images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png'
-    import figure1b from '../../../images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png'
-    import figure2 from '../../../images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency 1.png'
-    import figure2b from '../../../images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png'
-    import figure3 from '../../../images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png'
-    import figure3b from '../../../images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png'
-    import figure4 from '../../../images/blogs/accelerating-llama-2/Figure4-LLaMA-2-70B-Model-Throughput.png'
+	import figure1 from '../../../images/blogs/accelerating-llama-2/Figure1-LLaMA-2-7B-E2E-Throughput.png';
+	import figure1b from '../../../images/blogs/accelerating-llama-2/Figure1-LLaMA-2-13B-E2E-Throughput.png';
+	import figure2 from '../../../images/blogs/accelerating-llama-2/Figure2-LLaMA-2-7B-Prompt-Latency.png';
+	import figure2b from '../../../images/blogs/accelerating-llama-2/Figure2-LLaMA-2-13B-Prompt-Latency.png';
+	import figure3 from '../../../images/blogs/accelerating-llama-2/Figure3-LLaMA-2-7B-Tokens-Generated-Throughput.png';
+	import figure3b from '../../../images/blogs/accelerating-llama-2/Figure3-LLaMA-2-13B-Tokens-Generated-Throughput.png';
+	import figure4 from '../../../images/blogs/accelerating-llama-2/Figure4-LLaMA-2-70B-Model-Throughput.png';
 	import figure5 from '../../../images/blogs/accelerating-llama-2/LLaMA-2OptimizationDiagram-5.png';
 	import figure6 from '../../../images/blogs/accelerating-llama-2/LLaMAWindowsExportRotaryEmbeddingSubgraph-6.jpg';
 	import figure7 from '../../../images/blogs/accelerating-llama-2/RotaryEmbeddingFunctionExample-7.png';
@@ -18,12 +18,40 @@
 		name="description"
 		content="Explore how ONNX Runtime can propel your Llama2 variants for faster inference."
 	/>
+	<meta
+		name="keywords"
+		content="Accelerating LLaMA-2, ONNX Runtime, Inference, AI, Microsoft, Meta, Llama2, Performance Optimization, Multi-GPU Inference"
+	/>
+	<meta name="author" content="Kunal Vaishnavi, Parinita Rahi" />
+	<meta name="date" content="2023-11-14" />
+	<meta property="og:title" content="Accelerating LLaMA-2 Inference with ONNX Runtime" />
+	<meta
+		property="og:description"
+		content="Explore how ONNX Runtime accelerates LLaMA-2 inference, achieving up to 3.8X faster performance for models ranging from 7B to 70B parameters. Learn about graph fusions, kernel optimizations, multi-GPU inference support, and more."
+	/>
+	<meta property="og:type" content="article" />
+	<meta property="og:url" content="https://onnxruntime.ai/blogs/accelerating-llama-2" />
+	<meta property="og:image" content={figure5} />
+	<meta property="og:site_name" content="ONNX Runtime" />
+	<meta name="twitter:card" content={figure5} />
+	<meta name="twitter:title" content="Accelerating LLaMA-2 Inference with ONNX Runtime" />
+	<meta
+		name="twitter:description"
+		content="Explore how ONNX Runtime can propel your Llama2 variants for faster inference."
+	/>
+	<meta name="twitter:image" content={figure5} />
 </svelte:head>
 <Header pathvar="" />
 <div class="container mx-auto px-4 md:px-8 lg:px-48 pt-8">
 	<h1 class="text-5xl pb-2">Accelerating LLaMA-2 Inference with ONNX Runtime</h1>
-	<p class="text-neutral">By: <a href="https://www.linkedin.com/in/kunal-v-16315b94" class="text-blue-500">Kunal Vaishnavi</a> and <a href="https://www.linkedin.com/in/parinitaparinita/" class="text-blue-500">Parinita Rahi</a> </p>
-	<p class="text-neutral">14TH NOVEMBER, 2023</p>
+	<p class="text-neutral">
+		By: <a href="https://www.linkedin.com/in/kunal-v-16315b94" class="text-blue-500"
+			>Kunal Vaishnavi</a
+		>
+		and
+		<a href="https://www.linkedin.com/in/parinitaparinita/" class="text-blue-500">Parinita Rahi</a>
+	</p>
+	<p class="text-neutral">14TH NOVEMBER, 2023 <span class="italic text-stone-500">(Updated 22nd November)</span></p>
 	<div class="py-4">
 		<p class="mb-4">
 			Interested in running Llama2 faster? Let us explore how ONNX Runtime can propel your Llama2
@@ -31,7 +59,7 @@
 		</p>
 
 		<p class="mb-4">
-			You can now experience significant inference gains—up to 4X faster—for the 7B, 13B, and 70B
+			You can now experience significant inference gains—up to 3.8X faster—for the 7B, 13B, and 70B
 			models, thanks to state-of-the-art fusion and kernel optimizations with ONNX Runtime. This
 			blog details performance enhancements, dives into ONNX Runtime fusion optimizations, multi-GPU
 			inferencing support, and guides you on how to leverage the cross-platform prowess of ONNX
@@ -44,11 +72,13 @@
 
 		<p class="mb-4">
 			Llama2 is a state-of-the-art open source LLM from Meta ranging in scale from 7B to 70B
-			parameters (7B, 13B, 70B). Microsoft and Meta <a href="https://blogs.microsoft.com/blog/2023/07/18/microsoft-and-meta-expand-their-ai-partnership-with-llama-2-on-azure-and-windows/" class="text-blue-500">announced</a> their AI on Azure and Windows
-			collaboration in July 2023. As part of the announcement, Llama2 was added to the Azure AI
-			model catalog, which serves as a hub of foundation models that empower developers and machine
-			learning (ML) professionals to easily discover, evaluate, customize, and deploy pre-built
-			large AI models at scale.
+			parameters (7B, 13B, 70B). Microsoft and Meta <a
+				href="https://blogs.microsoft.com/blog/2023/07/18/microsoft-and-meta-expand-their-ai-partnership-with-llama-2-on-azure-and-windows/"
+				class="text-blue-500">announced</a
+			> their AI on Azure and Windows collaboration in July 2023. As part of the announcement, Llama2
+			was added to the Azure AI model catalog, which serves as a hub of foundation models that empower
+			developers and machine learning (ML) professionals to easily discover, evaluate, customize, and
+			deploy pre-built large AI models at scale.
 		</p>
 
 		<p class="mb-4">
@@ -65,95 +95,98 @@
 			As part of the new 1.16.2 release, ONNX Runtime now has several built-in optimizations for
 			Llama2, including graph fusions and kernel optimizations. The inference speedups, when
 			compared to Hugging Face (HF) variants of Llama2 in PyTorch compile mode for prompt latency of
-			CUDA FP16, are mentioned below. We see ~3X gains in end-to-end throughput comparisons for both
-			7B and 13B models. The end-to-end throughput or wall-clock throughput shown below is defined
-			as <i>batch size * (prompt length + token generation length) / wall-clock latency</i> where
-			wall-clock latency = the latency from running end-to-end and token generation length = 256
-			generated tokens. The E2E throughput is up to 4.5X more when compared to PyTorch compile.
+			CUDA FP16, are mentioned below. The end-to-end throughput or wall-clock throughput shown below
+			is defined as <i
+				>batch size * (prompt length + token generation length) / wall-clock latency</i
+			> where wall-clock latency = the latency from running end-to-end and token generation length =
+			256 generated tokens. The E2E throughput is 2.4X more (13B) and 1.8X more (7B) when compared to
+			PyTorch compile. For higher batch size, sequence length like 16, 2048 pytorch eager times out,
+			while ORT shows better performance than compile mode.
 		</p>
-        <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-            <figure class="px-10 pt-4">
-                <img src={figure1} alt="E2E Throughput Comparisons - Llama-2-7b" />
-            </figure>
-            <figure class="px-10 pt-4 my-auto">
-                <img src={figure1b} alt="E2E Throughput Comparisons - Llama-2-13b" />
-            </figure>
-        </div>
-        <div class="mt-2 mb-4 text-center">
-            Figure 1: E2E Throughput Comparisons
-        </div>
+		<div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
+			<figure class="px-10 pt-4">
+				<img src={figure1} alt="E2E Throughput Comparisons - Llama-2-7b" />
+			</figure>
+			<figure class="px-10 pt-4 my-auto">
+				<img src={figure1b} alt="E2E Throughput Comparisons - Llama-2-13b" />
+			</figure>
+		</div>
+		<div class="mt-2 mb-4 text-center">Figure 1: E2E Throughput Comparisons</div>
 
 		<h2 class="text-blue-500 text-3xl mb-4">Latency and Throughput</h2>
 
 		<p class="mb-4">
-			The graphs below show latency comparisons between the ONNX Runtime and PyTorch variants of the Llama2
-			7B model on CUDA FP16. Latency here is defined as the time it takes to complete one pass
-			through the model to produce the logits and synchronize the outputs.
+			The graphs below show latency comparisons between the ONNX Runtime and PyTorch variants of the
+			Llama2 7B model on CUDA FP16. Latency here is defined as the time it takes to complete one
+			pass through the model to produce the logits and synchronize the outputs.
 		</p>
-        
-        <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-            <figure class="px-10 pt-4">
-                <img src={figure2} alt="Prompt Latency Comparisons - Llama-2-7b" />
-            </figure>
-            <figure class="px-10 pt-4 my-auto">
-                <img src={figure2b} alt="Prompt Latency Comparisons - Llama-2-13b" />
-            </figure>
-        </div>
-        <div class="mt-2 mb-4 text-center">
-            Figure 2: Prompt Latency Comparisons
-        </div>
+
+		<div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
+			<figure class="px-10 pt-4">
+				<img src={figure2} alt="Prompt Latency Comparisons - Llama-2-7b" />
+			</figure>
+			<figure class="px-10 pt-4 my-auto">
+				<img src={figure2b} alt="Prompt Latency Comparisons - Llama-2-13b" />
+			</figure>
+		</div>
+		<div class="mt-2 mb-4 text-center">Figure 2: Prompt Latency Comparisons</div>
 
 		<p class="mb-4">
-			Token generation throughput below is the average throughput of the first 128 tokens generated.
-			We see up to 3.5X gains in token generation throughput when compared to PyTorch eager and
-			compile modes.
+			Token generation throughput below is the average throughput of the first 256 tokens generated.
+			We see up to ~1.4X (7B) and ~1.7X (13B) gains in token generation throughput when compared to
+			PyTorch compile mode.
 		</p>
 
-        <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-            <figure class="px-10 pt-4">
-                <img src={figure3} alt="Tokens Generated Throughput Comparisons - Llama-2-7b" />
-            </figure>
-            <figure class="px-10 pt-4 my-auto">
-                <img src={figure3b} alt="Tokens Generated Throughput Comparisons - Llama-2-13b" />
-            </figure>
-        </div>
-        <div class="mt-2 mb-4 text-center">
-            Figure 3: Tokens Generated Throughput Comparisons
-        </div>
+		<div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
+			<figure class="px-10 pt-4">
+				<img src={figure3} alt="Tokens Generated Throughput Comparisons - Llama-2-7b" />
+			</figure>
+			<figure class="px-10 pt-4 my-auto">
+				<img src={figure3b} alt="Tokens Generated Throughput Comparisons - Llama-2-13b" />
+			</figure>
+		</div>
+		<div class="mt-2 mb-4 text-center">Figure 3: Tokens Generated Throughput Comparisons</div>
 
-		<p class="mb-4">More details on these metrics can be found <a href="https://github.com/microsoft/onnxruntime-inference-examples/blob/main/python/models/llama2/README.md" class="text-blue-500">here</a>.</p>
+		<p class="mb-4">
+			More details on these metrics can be found <a
+				href="https://github.com/microsoft/onnxruntime-inference-examples/blob/main/python/models/llama2/README.md"
+				class="text-blue-500">here</a
+			>.
+		</p>
 
 		<h2 class="text-blue-500 text-3xl mb-4">ONNX Runtime with Multi-GPU Inference</h2>
 
 		<p class="mb-4">
-			ONNX Runtime supports multi-GPU inference to enable serving large models. Even in FP16 precision,
-			the LLaMA-2 70B model requires 140GB. Loading the model requires multiple GPUs 
-			for inference, even with a powerful NVIDIA A100 80GB GPU.
+			ONNX Runtime supports multi-GPU inference to enable serving large models. Even in FP16
+			precision, the LLaMA-2 70B model requires 140GB. Loading the model requires multiple GPUs for
+			inference, even with a powerful NVIDIA A100 80GB GPU.
 		</p>
 
 		<p class="mb-4">
-			ONNX Runtime applied <a href="https://arxiv.org/pdf/1909.08053.pdf" class="text-blue-500">Megatron-LM</a> Tensor Parallelism on the 70B model to split the
-			original model weight onto different GPUs. Megatron sharding on the 70B model
-			shards the PyTorch model with FP16 precision into 4 partitions, converts each partition into ONNX
-			format, and then applies a new ONNX Runtime graph fusion on the converted ONNX model. The 70B
-			model has ~30 tokens per second throughput for token generation at batch size 1, and
-			end-to-end throughput starts at 30 ms for smaller sequence lengths with these optimizations.
-			You can find additional example scripts <a href="https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/models/llama/" class="text-blue-500">here</a>.
+			ONNX Runtime applied <a href="https://arxiv.org/pdf/1909.08053.pdf" class="text-blue-500"
+				>Megatron-LM</a
+			>
+			Tensor Parallelism on the 70B model to split the original model weight onto different GPUs. Megatron
+			sharding on the 70B model shards the PyTorch model with FP16 precision into 4 partitions, converts
+			each partition into ONNX format, and then applies a new ONNX Runtime graph fusion on the converted
+			ONNX model. The 70B model has ~30 tokens per second throughput for token generation at batch size
+			1, and end-to-end throughput starts at 30 tps for smaller sequence lengths with these optimizations.
+			You can find additional example scripts
+			<a
+				href="https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/models/llama/"
+				class="text-blue-500">here</a
+			>.
 		</p>
-        
-        <figure class="px-10 pt-4">
-			<img src={figure4} alt="70B Llama2 Model Throughput" class="w-3/5 mx-auto"/>
-			<figcaption class="mt-2 mb-4 text-center">
-				Figure 4: 70B Llama2 Model Throughput
-			</figcaption>
+
+		<figure class="px-10 pt-4">
+			<img src={figure4} alt="70B Llama2 Model Throughput" class="w-3/5 mx-auto" />
+			<figcaption class="mt-2 mb-4 text-center">Figure 4: 70B Llama2 Model Throughput</figcaption>
 		</figure>
 
 		<h2 class="text-blue-500 text-3xl mb-4">ONNX Runtime Optimizations</h2>
 		<figure class="px-10 pt-4">
 			<img src={figure5} alt="LLaMA-2 Optimization Diagram" />
-			<figcaption class="mt-2 mb-4 text-center">
-				Figure 5: LLaMA-2 Optimization Diagram
-			</figcaption>
+			<figcaption class="mt-2 mb-4 text-center">Figure 5: LLaMA-2 Optimization Diagram</figcaption>
 		</figure>
 
 		<p class="mb-4">
@@ -165,7 +198,7 @@
 			fusion opportunities can instead be identified by exporting a large module as a function and
 			then pattern matching against a function's spec.
 		</p>
-        <figure class="px-10 pt-4">
+		<figure class="px-10 pt-4">
 			<img src={figure6} alt="Example of Rotary Embedding Function" />
 			<figcaption class="mt-2 mb-4 text-center">
 				Figure 6: Example of Rotary Embedding Function
@@ -179,7 +212,7 @@
 			the inputs and outputs and represent all these nodes as a single operator.
 		</p>
 
-        <figure class="px-10 pt-4">
+		<figure class="px-10 pt-4">
 			<img src={figure7} alt="Example of Rotary Embedding Function in Parent Graph" />
 			<figcaption class="mt-2 mb-4 text-center">
 				Figure 7: Example of Rotary Embedding Function in Parent Graph
@@ -195,11 +228,11 @@
 		</p>
 
 		<p class="mb-4">
-			ONNX Runtime also adds support for the GroupQueryAttention (GQA) operator, which leverages the new
-			Flash Attention V2 algorithm and its optimized kernels to efficiently compute attention. The
-			GQA operator supports past-present buffer sharing between the past key/value cache (past KV
-			cache) and the present key/value cache (present KV cache). By binding the present KV caches to
-			the past KV caches, there is no need to allocate separate on-device memory for both caches.
+			ONNX Runtime also adds support for the GroupQueryAttention (GQA) operator, which leverages the
+			new Flash Attention V2 algorithm and its optimized kernels to efficiently compute attention.
+			The GQA operator supports past-present buffer sharing between the past key/value cache (past
+			KV cache) and the present key/value cache (present KV cache). By binding the present KV caches
+			to the past KV caches, there is no need to allocate separate on-device memory for both caches.
 			Instead, the past KV caches can be pre-allocated with enough on-device memory so that no new
 			on-device memory needs to be requested during inference. This reduces memory usage when the KV
 			caches become large during compute-intensive workloads and lowers latency by eliminating
@@ -209,20 +242,29 @@
 		</p>
 
 		<p class="mb-4">
-			In addition to these fusions and kernel optimizations, ONNX Runtime reduces the model’s memory usage. 
-			Besides quantization improvements (which will be covered in a future post), ONNX Runtime compresses the 
-			size of the cosine and sine caches used in each of the rotary embeddings by 50%. The compute kernels in 
-			ONNX Runtime that run the rotary embedding computations can then recognize this format and use their 
-			parallelized implementations to calculate the rotary embeddings more efficiently with less memory usage. 
-			The rotary embedding compute kernels also support interleaved and non-interleaved formats to support both 
-			the <a href="https://github.com/microsoft/Llama-2-Onnx" class="text-blue-500">Microsoft version of LLaMA-2</a>
+			In addition to these fusions and kernel optimizations, ONNX Runtime reduces the model’s memory
+			usage. Besides quantization improvements (which will be covered in a future post), ONNX
+			Runtime compresses the size of the cosine and sine caches used in each of the rotary
+			embeddings by 50%. The compute kernels in ONNX Runtime that run the rotary embedding
+			computations can then recognize this format and use their parallelized implementations to
+			calculate the rotary embeddings more efficiently with less memory usage. The rotary embedding
+			compute kernels also support interleaved and non-interleaved formats to support both the <a
+				href="https://github.com/microsoft/Llama-2-Onnx"
+				class="text-blue-500">Microsoft version of LLaMA-2</a
+			>
 			and the Hugging Face version of LLaMA-2 respectively while sharing the same calculations.
 		</p>
 
 		<p class="mb-4">
-			The optimizations work for the <a href="https://huggingface.co/meta-llama" class="text-blue-500">Hugging Face versions</a> (models ending with <i>-hf</i>) and the Microsoft versions. 
-			You can download the optimized HF versions from <a href="https://github.com/microsoft/Llama-2-Onnx/tree/main-CUDA_CPU" class="text-blue-500">Microsoft's LLaMA-2 ONNX repository</a>. Stay tuned for
-			newer Microsoft versions coming soon!
+			The optimizations work for the <a
+				href="https://huggingface.co/meta-llama"
+				class="text-blue-500">Hugging Face versions</a
+			>
+			(models ending with <i>-hf</i>) and the Microsoft versions. You can download the optimized HF
+			versions from
+			<a href="https://github.com/microsoft/Llama-2-Onnx/tree/main-CUDA_CPU" class="text-blue-500"
+				>Microsoft's LLaMA-2 ONNX repository</a
+			>. Stay tuned for newer Microsoft versions coming soon!
 		</p>
 
 		<h2 class="text-blue-500 text-3xl mb-4">Optimize your own model using Olive</h2>
@@ -235,19 +277,24 @@
 		</p>
 
 		<p class="mb-4">
-			Here is an example of <a href="https://github.com/microsoft/Olive/tree/main/examples/llama2" class="text-blue-500">Llama2 optimization with Olive</a>, which harnesses ONNX Runtime
-			optimizations highlighted in this blog. Distinct optimization flows cater to various
-			requirements. For instance, you have the flexibility to choose different data types for
-			quantization in CPU and GPU inference, based on your accuracy tolerance. Additionally, you can
-			fine-tune your own Llama2 model with Olive-QLoRa on client GPUs and perform inference with
-			ONNX Runtime optimizations.
+			Here is an example of <a
+				href="https://github.com/microsoft/Olive/tree/main/examples/llama2"
+				class="text-blue-500">Llama2 optimization with Olive</a
+			>, which harnesses ONNX Runtime optimizations highlighted in this blog. Distinct optimization
+			flows cater to various requirements. For instance, you have the flexibility to choose
+			different data types for quantization in CPU and GPU inference, based on your accuracy
+			tolerance. Additionally, you can fine-tune your own Llama2 model with Olive-QLoRa on client
+			GPUs and perform inference with ONNX Runtime optimizations.
 		</p>
 
 		<h2 class="text-blue-500 text-3xl mb-4">Usage Example</h2>
 
 		<p class="mb-4">
-			Here is a <a href="https://github.com/microsoft/onnxruntime-inference-examples/blob/main/python/models/llama2/LLaMA-2%20E2E%20Notebook.ipynb" class="text-blue-500">sample notebook</a> that shows you an end-to-end example of how you can use the above
-			ONNX Runtime optimizations in your application.
+			Here is a <a
+				href="https://github.com/microsoft/onnxruntime-inference-examples/blob/main/python/models/llama2/LLaMA-2%20E2E%20Notebook.ipynb"
+				class="text-blue-500">sample notebook</a
+			> that shows you an end-to-end example of how you can use the above ONNX Runtime optimizations
+			in your application.
 		</p>
 
 		<h2 class="text-blue-500 text-3xl mb-4">Conclusion</h2>
diff --git a/src/routes/blogs/pytorch-on-the-edge/+page.svelte b/src/routes/blogs/pytorch-on-the-edge/+page.svelte
index d140bc8d7c..6d7f950f51 100644
--- a/src/routes/blogs/pytorch-on-the-edge/+page.svelte
+++ b/src/routes/blogs/pytorch-on-the-edge/+page.svelte
@@ -150,10 +150,41 @@ fun run(audioTensor: OnnxTensor): Result {
 		name="description"
 		content="Everything you need to know about running PyTorch models on the edge with ONNX Runtime."
 	/>
+	<meta name="title" content="Run PyTorch models on the edge" />
+	<meta
+		name="keywords"
+		content="PyTorch, ONNX Runtime, edge computing, machine learning, deep learning, model optimization, model deployment, AI on edge"
+	/>
+	<meta name="author" content="Natalie Kershaw, Prasanth Pulavarthi" />
+	<meta name="date" content="2023-10-12" />
+	<meta name="image" content={ORT} />
+	<meta name="robots" content="index, follow" />
+	<meta name="og:title" content="Run PyTorch models on the edge" />
+	<meta
+		name="og:description"
+		content="Learn how to run PyTorch models on edge devices using ONNX Runtime. Explore considerations, tools, and examples for deploying PyTorch models on Windows, in the browser, on mobile, and for on-device training."
+	/>
+	<meta name="og:type" content="article" />
+	<meta name="og:url" content="https://onnxruntime.ai/blogs/pytorch-on-the-edge" />
+	<meta name="og:image" content={ORT} />
+	<meta name="twitter:title" content="Run PyTorch models on the edge" />
+	<meta
+		name="twitter:description"
+		content="Learn how to run PyTorch models on edge devices using ONNX Runtime. Explore considerations, tools, and examples for deploying PyTorch models on Windows, in the browser, on mobile, and for on-device training."
+	/>
+	<meta name="twitter:card" content={ORT} />
+	<meta name="twitter:image" content={ORT} />
 </svelte:head>
 <Header pathvar="" />
 <div class="container mx-auto px-4 md:px-8 lg:px-48 pt-8">
 	<h1 class="text-5xl pb-2">Run PyTorch models on the edge</h1>
+	<p class="text-neutral">
+		By: <a href="https://www.linkedin.com/in/natkershaw/" class="text-blue-500">Natalie Kershaw</a>
+		and
+		<a href="https://www.linkedin.com/in/prasanthpulavarthi/" class="text-blue-500"
+			>Prasanth Pulavarthi</a
+		>
+	</p>
 	<p class="text-neutral">12TH OCTOBER, 2023</p>
 	<div class="py-4">
 		<div class="col-span-12 md:col-span-9">
diff --git a/src/routes/components/code-blocks.svelte b/src/routes/components/code-blocks.svelte
index 32acbe0b70..e5e1793ef4 100644
--- a/src/routes/components/code-blocks.svelte
+++ b/src/routes/components/code-blocks.svelte
@@ -68,7 +68,7 @@
 				>
 			</div>
 		</div>
-		<div class="hidden lg:block col-span-2 mx-auto tab-container">
+		<div class="hidden lg:block col-span-2 mx-auto min-w-[675px] min-h-[400px]">
 			<div class="tabs">
 				<p
 					on:mouseenter={handleClick}
@@ -160,10 +160,3 @@
 	</div>
 	<div class="divider px-6" />
 </div>
-
-<style>
-	.tab-container {
-		min-width: 675px;
-		min-height: 400px;
-	}
-</style>
diff --git a/src/routes/getting-started/table.svelte b/src/routes/getting-started/table.svelte
index 8bbed8f0a3..62ba558645 100644
--- a/src/routes/getting-started/table.svelte
+++ b/src/routes/getting-started/table.svelte
@@ -669,7 +669,7 @@
 			"Follow build instructions from <a class='text-blue-500' href='http://www.onnxruntime.ai/docs/execution-providers/community-maintained/CANN-ExecutionProvider.html#build' target='_blank'>here</a>.",
 
 		'linux,Python,X64,CANN':
-			"pip install onnxruntime-cann <br/>Refer to <a class='text-blue-500' href='http://www.onnxruntime.ai/docs/execution-providers/community-maintained/community-maintained/CANN-ExecutionProvider.html#requirements' target='_blank'>docs</a> for requirements.",
+			"pip install onnxruntime-cann <br/>Refer to <a class='text-blue-500' href='http://www.onnxruntime.ai/docs/execution-providers/community-maintained/CANN-ExecutionProvider.html#requirements' target='_blank'>docs</a> for requirements.",
 
 		'linux,C-API,X64,CANN':
 			"Follow build instructions from <a class='text-blue-500' href='http://www.onnxruntime.ai/docs/execution-providers/community-maintained/CANN-ExecutionProvider.html#build' target='_blank'>here</a>.",