Update README.md

README.md

@@ -9,10 +9,11 @@ inference: false
 This repository hosts the optimized versions of [Phi-3.5-mini-4k-instruct](https://aka.ms/phi3.5-mini-4k-instruct) to accelerate inference with ONNX Runtime.
 Optimized Phi-3.5 Mini models are published here in [ONNX](https://onnx.ai) format to run with [ONNX Runtime](https://onnxruntime.ai/) on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets.
 
-To easily get started with Phi-3.5, you can use our newly introduced ONNX Runtime Generate() API. See See [here](https://aka.ms/generate-tutorial) for instructions on how to run it.
+To easily get started with Phi-3.5, you can use our newly introduced ONNX Runtime Generate() API. See [here](https://aka.ms/generate-tutorial) for instructions on how to run it.
 
 ## ONNX Models 
 Here are some of the optimized configurations we have added:
+
 1.	ONNX model for fp16 CUDA: ONNX model you can use to run for your NVIDIA GPUs.
 2.	ONNX model for int4 CUDA: ONNX model for NVIDIA GPUs using int4 quantization via AWQ.
 3.	ONNX model for int4 CPU and Mobile: ONNX model for CPU and mobile using int4 quantization via AWQ.
@@ -21,31 +22,38 @@ Here are some of the optimized configurations we have added:
 Phi-3.5-mini is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data. The model belongs to the Phi-3 model family and supports 128K token context length. The model underwent a rigorous enhancement process, incorporating both supervised fine-tuning, proximal policy optimization, and direct preference optimization to ensure precise instruction adherence and robust safety measures.
 
 ## Intended Uses
-Primary Use Cases
 The Phi 3.5 model is intended for commercial and research use in multiple languages. The model provides uses for general purpose AI systems and applications which require:
+
 1.	Memory/compute constrained environments
 2.	Latency bound scenarios
 3.	Strong reasoning (especially code, math and logic)
+
 ## Use Case Considerations
 Phi 3.5 models are not specifically designed or evaluated for all downstream purposes. Developers should consider common limitations of language models as they select use cases, and evaluate and mitigate for accuracy, safety, and fariness before using within a specific downstream use case, particularly for high risk scenarios. Developers should be aware of and adhere to applicable laws or regulations (including privacy, trade compliance laws, etc.) that are relevant to their use case.
 Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.
-Release Notes
-This is an update over the June 2024 instruction-tuned Phi-3 Mini ONNX release based. We believe most use cases will benefit from this release, but we encourage users to test their particular AI applications. We appreciate the enthusiastic adoption of the Phi-3 model family and continue to welcome all feedback from the community.
+
+## Release Notes
+This is an update over the instruction-tuned Phi-3 Mini ONNX model release. We believe most use cases will benefit from this release, but we encourage users to test their particular AI applications. We appreciate the enthusiastic adoption of the Phi-3 model family and continue to welcome all feedback from the community.
+
 ## Hardware Supported
 The ONNX models are tested on:
+
 - GPU SKU: RTX 4090 (DirectML)
 - GPU SKU: 1 A100 80GB GPU, SKU: Standard_ND96amsr_A100_v4 (CUDA)
-- CPU SKU: Standard D16s v6 (16 vcpus, 64 GiB memory) , AMD CPU: Internal_D64as_v5
+- CPU SKU: Standard D16s v6 (16 vcpus, 64 GiB memory)
+- AMD CPU: Internal_D64as_v5
+
+Minimum Configuration Required:
+- Windows: DirectX 12-capable GPU and a minimum of 4GB of combined RAM
+- CUDA: NVIDIA GPU with Compute Capability >= 7.0
 
-## Minimum Configuration Required:
-•	Windows: DirectX 12-capable GPU and a minimum of 4GB of combined RAM
-•	CUDA: NVIDIA GPU with Compute Capability >= 7.0
 ## Model Description
-•	Developed by: Microsoft
-•	Model type: ONNX
-•	Language(s) (NLP): Python, C, C++
-•	License: MIT
-•	Model Description: This is a conversion of the Phi-3.5 Mini-Instruct model for ONNX Runtime inference.
+
+- **Developed by:**  Microsoft
+- **Model type:** ONNX
+- **Language(s) (NLP):** Python, C, C++
+- **License:** MIT
+- **Model Description:** This is a conversion of the Phi-3.5 Mini-Instruct model for ONNX Runtime inference.
 
 ## How to Get Started with the Model
 To make running of the Phi-3 models across a range of devices and platforms across various execution provider backends possible, we introduce a new API to wrap several aspects of generative AI inferencing. This API make it easy to drag and drop LLMs straight into your app. For running the early version of these models with ONNX Runtime, follow the steps [here](http://aka.ms/generate-tutorial).
@@ -53,7 +61,7 @@ To make running of the Phi-3 models across a range of devices and platforms acro
 For example:
 
 ```python
-python model-qa.py -m /*{YourModelPath}*/onnx/cpu_and_mobile/phi-3-mini-4k-instruct-int4-cpu -k 40 -p 0.95 -t 0.8 -r 1.0
+python model-qa.py -m /*{YourModelPath}*/Phi-3.5-mini-instruct-onnx/cpu_and_mobile/cpu-int4-awq-block-128-acc-level-4 -k 40 -p 0.95 -t 0.8 -r 1.0
 ```
 
 ```
@@ -66,10 +74,10 @@ This joke plays on the double meaning of "make up." In science, atoms are the fu
 ```
 
 ## Performance Metrics
-CUDA
+
 Phi-3.5 Mini-Instruct performs better in ONNX Runtime than PyTorch for all batch size, prompt length combinations. 
-The table below shows the average throughput of the first 256 tokens generated (tps) for FP16 and INT4 precisions on CUDA as measured on 1 A100 80GB GPU, SKU: Standard_ND96amsr_A100_v4. ONNX Runtime Models for GPU are 21X faster than PyTorch Compile and up to 8X faster than llama.cpp on A100 GPU.
-	Average Throughput of First 256 Tokens Generated (tps)	
+
+The table below shows the average throughput of the first 256 tokens generated (tps) for FP16 and INT4 precisions on CUDA as measured on 1 A100 80GB GPU, SKU: Standard_ND96amsr_A100_v4. ONNX Runtime models for GPU are 21X faster than PyTorch Compile and up to 8X faster than llama.cpp on A100 GPU.
 
 | Batch Size, Sequence Length | ONNX RT INT4 | PyTorch Eager INT4 | PyTorch Compile INT4 | Llama.cpp INT4 | INT4 SpeedUp ORT/PyTorch Eager | INT4 SpeedUp ORT/PyTorch Compile | INT4 SpeedUp ORT/Llama.cpp |
 | --- | --- | --- | --- | --- | --- | --- | --- |
@@ -97,7 +105,8 @@ The table below shows the average throughput of the first 256 tokens generated (
 | 16,1024 | 536.73 | 209.13 | 169.30 | 71.57 | 2.57 | 3.17 | 7.50 |
 | 16,2048 | 375.31 | 153.95 | 158.77 | 45.97 | 2.44 | 2.36 | 8.16 |
 | 16,3840 | 243.66 | OOM | OOM | 28.33 |  |  | 8.60 |
-							
+
+                            
 | Batch Size, Sequence Length | ONNX RT FP16 | PyTorch Eager FP16 | PyTorch Compile FP16 | Llama.cpp | FP16 SpeedUp ORT/PyTorch Eager | FP16 SpeedUp ORT/PyTorch Compile | FP16 SpeedUp ORT/Llama.cpp |
 | --- | --- | --- | --- | --- | --- | --- | --- |
 | 1,16 | 137.30 | 26.02 | 26.83 | 125.86 | 5.28 | 5.12 | 1.09 |
@@ -125,8 +134,9 @@ The table below shows the average throughput of the first 256 tokens generated (
 | 16,2048 | 441.15 | 121.17 | 162.93 | 41.30 | 3.64 | 2.71 | 10.68 |
 | 16,3840 | 270.38 | OOM | OOM | 26.50 | 0.00 | 0.00 | 10.20 |
 
+
 The table below shows the average throughput of the first 256 tokens generated (tps) for INT4 precision on CPU as measured on a Standard D16s v6 (16 vcpus, 64 GiB memory)
-	Average Throughput of First 256 Tokens Generated (tps)
+
 | Batch Size, Sequence Length | ORT INT4 AWQ | Llama.cpp INT4 | INT4 AWQ SpeedUp Llama.cpp |
 | --- | --- | --- | --- |
 | 1,16 | 41.99 | 26.72 | 1.57 |
@@ -134,24 +144,33 @@ The table below shows the average throughput of the first 256 tokens generated (
 | 1,256 | 41.26 | 26.30 | 1.57 |
 | 1,1024 | 37.15 | 24.02 | 1.55 |
 | 1,2048 | 32.68 | 21.82 | 1.50 |
-		
+
+        
 ## Package Versions
-Pip packages	torch	2.4.1
-	triton	3.0.0
-	onnxruntime	1.19.2
-	transformers	4.44.2
-llama.cpp	bdf314f38a2c90e18285f7d7067e8d736a14000a		
+
+| Pip package name           | Version  |
+|----------------------------|----------|
+| torch                      | 2.4.1    |
+| triton                     | 3.0.0    |
+| onnxruntime-gpu            | 1.19.2   |
+| onnxruntime-genai          | 0.4.0    |
+| onnxruntime-genai-cuda     | 0.4.0    |
+| transformers               | 4.44.2   |
+| llama.cpp                  | bdf314f38a2c90e18285f7d7067e8d736a14000a   |
+
 
 ## Appendix
-Activation Aware Quantization" AWQ works by identifying the top 1% most salient weights that are most important for maintaining accuracy and quantizing the remaining 99% of weights. This leads to less accuracy loss from quantization compared to many other quantization techniques. For more on AWQ, see here.
+
+Activation Aware Quantization (AWQ) works by identifying the top 1% most salient weights that are most important for maintaining accuracy and quantizing the remaining 99% of weights. This leads to less accuracy loss from quantization compared to many other quantization techniques. For more on AWQ, see [here](https://arxiv.org/abs/2306.00978).
 
 ## Model Card Contact
 parinitarahi 
+
 ## Contributors
- Sunghoon Choi, Yufeng Li, Kunal Vaishnavi, Akshay Sonawane,  Rui Ren, Parinita Rahi
+Sunghoon Choi, Yufeng Li, Kunal Vaishnavi, Akshay Sonawane, Rui Ren, Parinita Rahi
+
 ## License
 The model is licensed under the MIT license.
-## Trademarks
-This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft’s Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party’s policies.
-
 
+## Trademarks
+This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft’s Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party’s policies.