Add pre-generated prompts for benchmark

Author: omer-demir

benchmark/python/benchmark_e2e.py

@@ -83,6 +83,14 @@ def generate_prompt(model, tokenizer, prompt_length, use_graph_capture) -> str:
         generator.generate_next_token()
     return tokenizer.decode(generator.get_sequence(0))
 
+# Use prompt length to get pre-defined prompt
+def get_prompt_by_length(prompt_length):
+    json_path = "prompts.json"
+    with open(json_path) as prompts_file:
+        content = prompts_file.read()
+        data = json.load(content)
+        return data[f"{prompt_length}"]
+
 def get_target_pip_package_version(target_pip_package_name_list):
     # get package name and version
     import pkg_resources
@@ -232,6 +240,9 @@ def run_benchmark(args, batch_size, prompt_length, generation_length, max_length
         # use random tokens instead of generating a prompt using the model and then tokenizing it
         tokens = np.random.randint(100, size=(batch_size, prompt_length))
         prompt = [tokenizer.decode(tokens[0])] * batch_size
+    elif args.use_prompt_set:
+        prompt = get_prompt_by_length(prompt_length)
+        tokens = tokenizer.encode_batch(prompt)
     else:
         prompt = [generate_prompt(model, tokenizer, prompt_length, args.use_graph_capture)] * batch_size
         tokens = tokenizer.encode_batch(prompt)
@@ -424,6 +435,7 @@ def str2strlist(value):
     parser.add_argument('-mn', '--model_name', type=str, default='model_name', help='Model name defined by users')
     parser.add_argument('-pr', '--precision', type=str, default='fp16', help='Model precision for metrics info')
     parser.add_argument('--use_random_tokens', action='store_true', help='Use random tokens instead of generating a prompt')
+    parser.add_argument('--use_prompt_set', action='store_true', help='Use pre-generated prompt set instead of generating a prompt')
     args = parser.parse_args()
 
     # check max_lengths

benchmark/python/prompts.json

@@ -0,0 +1,7 @@
+{
+    "16": "How are astronauts launched into space quickly on those rockets? ",
+    "64": "Today, we will learn how to bake a chocolate cake. First, you need to have all of the ingredients to bake. Otherwise, the chocolate cake won't be tasty. You will also need a large baking pan to hold the batter. ",
+    "256": "Risk Management and Insurance (RMI) is a field that focuses on the identification, assessment and financial mitigation of risk. It's about insurance but also more than that. For example, insurance companies look at risk factors such as age, gender and medical history to determine how much they will charge for life insurance coverage. However, RMI is not just about buying insurance (although it is a big part of this). It is also about taking steps to reduce the likelihood that something bad happens in the first place. For example, you may think twice before crossing a busy road if there's a high risk of being hit by a car or getting injured. In addition to insurance companies and financial services firms, RMI professionals work with individuals (customers), businesses and other entities (clients). Their job is to identify potential risks and help mitigate them before they become problems for their clients. This can include helping people prepare financially for unexpected events like losing a job or being injured in an accident, as well as assisting businesses with managing risk exposure from things like natural disasters or cyber attacks. Insurance companies use RMI to ",
+    "1024": "Risk Management and Insurance (RMI) is a field that focuses on the identification, assessment and financial mitigation of risk. It's about insurance but also more than that. For example, insurance companies look at risk factors such as age, gender and medical history to determine how much they will charge for life insurance coverage. However, RMI is not just about buying insurance (although it is a big part of this). It is also about taking steps to reduce the likelihood that something bad happens in the first place. For example, you may think twice before crossing a busy road if there's a high risk of being hit by a car or getting injured. In addition to insurance companies and financial services firms, RMI professionals work with individuals (customers), businesses and other entities (clients). Their job is to identify potential risks and help mitigate them before they become problems for their clients. This can include helping people prepare financially for unexpected events like losing a job or being injured in an accident, as well as assisting businesses with managing risk exposure from things like natural disasters or cyber attacks. Insurance companies use RMI to assess the level of risk associated with potential customers and determine how much they should charge them for coverage. For example, if you are a healthy 25-year old male who doesn't smoke and has never been in an accident, your insurance premiums will likely be lower than those of someone else who fits into one or more of these categories (or all three). Risk Management & Insurance is the process by which you can protect yourself from financial loss. It's about taking control of your money and making sure that it's safe, secure and accessible to you when you need it most. The first step in risk management is understanding what risks are important to you as an individual or a family member who may depend on the income generated by these investments for their livelihood. Once you have identified these key risk factors, then we can help identify how best to manage them through various strategies such as setting up automatic payments into savings accounts so that money is always available when needed most; setting aside emergency funds in case something unexpected happens (e.g., illness); investing wisely so that returns outpace inflation over time; diversifying portfolios by adding stocks and bonds which will help reduce volatility while still providing growth potential through dividends/interest payments over longer periods of time than if invested solely into one type of asset class alone etc. The field of risk management and insurance is growing rapidly, as more people become aware of the potential dangers that can arise from an unforeseen event or accident. As a result, there are many different careers within this field that you may want to consider if you're interested in working with risks and helping others protect themselves from them.One common career path in risk management is as an insurance agent/broker. This person would work for an insurance company or brokerage firm, selling policies to clients who need coverage against things like car accidents or home damage caused by natural disasters such as fires or floods. Insurance agents typically work on commission (i.e., they receive a percentage of every sale). This is important because it means that the more successful an agent is at selling policies, the higher his/her income will be. Another career option within risk management is working for an insurance company itself rather than as an external broker or salesperson. In this case, you'd help manage claims made by policyholders who have been injured through no fault of their own (for example after being hit by another driver). You can also work in risk analysis, a field that involves analyzing the potential risks associated with various investments and projects. This is done to determine whether or not an opportunity has enough upside to justify taking on any related risks. In addition, you might also be responsible for developing strategies to minimize those risks so they don't result in big losses if something goes wrong down the road. If your goal is to work as a broker or agent, then there are some prerequisites that will need to be met before beginning this career path: You must have an associate's degree from an accredited college; pass an exam administered by state regulators (the Series 6) and/or complete additional training offered by professional organizations such as NAFA, which stands for National Association of Financial Advisors. After meeting these requirements, you'll then need to find employment at one or more insurance companies where they offer positions that allow new hires some flexibility when starting out their careers.Risk management and insurance is a broad field that includes many different types of jobs. ",
+    "2048": "Artificial Intelligence (AI) is a transformative technology that has the potential to revolutionize society in many ways. AI can be used to enhance the accuracy and efficiency of decision-making, improve lives through new apps and services, and solve some of the thorny policy problems of climate change, infrastructure, and healthcare. In this essay, I will discuss some of the ways AI can benefit society. One of the most significant benefits of AI is its ability to improve healthcare. AI can assist doctors, nurses, and other healthcare professionals in making better diagnoses and faster decisions on a course of treatment, based on the large amount of data that currently exists. AI allows doctors to pinpoint effective drugs that may have otherwise been overlooked and can identify higher-risk individuals before any human can. AI can also help relieve the burden on healthcare professionals by taking care of routine data collection and filing, freeing up time for other higher-value activities. Another area where AI can benefit society is in the fight against climate change. AI can be used to analyze vast amounts of data, identify patterns, and provide accurate predictions. It can help us forecast what further spread of pandemics is going to look like, and track their development around the world. AI can also help us predict the impact of climate change on our planet and develop strategies to mitigate its effects. For example, AI can be used to optimize energy consumption, reduce waste, and improve the efficiency of transportation systems. AI can also benefit society by improving education. AI-powered educational tools can help students learn more effectively by providing personalized learning experiences tailored to their individual needs. AI can also help teachers by automating routine tasks such as grading and providing feedback on student work. This can free up time for teachers to focus on more important tasks such as lesson planning and student engagement. AI can also benefit society by improving public safety. AI-powered surveillance systems can help law enforcement agencies detect and prevent crime more effectively. AI can also be used to analyze social media data to identify potential threats and prevent them before they occur. For example, AI can be used to detect hate speech and other forms of online harassment, which can help prevent cyberbullying and other forms of online abuse. Finally, AI can benefit society by improving the economy. AI can help businesses become more efficient by automating routine tasks and providing insights into customer behavior. This can help businesses make better decisions and improve their bottom line. AI can also help create new jobs by enabling the development of new products and services that were previously impossible. In conclusion, AI has the potential to benefit society in many ways. From improving healthcare and education to fighting climate change and improving public safety, AI can help us solve some of the most pressing problems facing our world today. As we continue to develop and refine this transformative technology, it is important that we do so in an ethical and responsible manner, ensuring that the benefits of AI are shared by all members of society. AI has been a topic of discussion for many years, and while it has brought many benefits to society, there are also concerns about its impact. In this essay, I will discuss some of the reasons why AI may not help society. Firstly, AI can be biased. AI systems are designed by humans, and they can be infused with the biases of their creators. This can lead to discrimination against certain groups of people and can perpetuate existing inequalities in society. Additionally, AI can lack transparency, making it difficult to understand how decisions are being made. This can lead to mistrust of AI systems and can hinder their adoption. Secondly, AI can be used to automate jobs, which can lead to unemployment. While AI can increase productivity and efficiency, it can also lead to job displacement, particularly in industries that rely heavily on manual labor. This can have a negative impact on individuals and communities, particularly those that are already marginalized. Thirdly, AI can be used to create fake content, such as deepfakes, which can be used to spread misinformation and propaganda. This can have serious consequences for democracy and can undermine trust in institutions. Fourthly, AI can be used to create autonomous weapons, which can have devastating consequences. These weapons can make decisions without human intervention, which can lead to unintended consequences and can be difficult to control. Fifthly, AI can be used to create surveillance systems that infringe on privacy rights. These systems can be used to monitor individuals without their knowledge or consent, which can have serious consequences for civil liberties. In conclusion, while AI has many potential benefits, there are also concerns about its impact on society. It is important to consider these concerns and to ensure that AI is developed and used in a responsible and ethical manner. Within AI, there are also many subfields. Reinforcement learning is a type of machine learning algorithm that focuses on training models to make decisions in an environment in order to maximize a reward. This is typically done through trial and error, as the algorithm receives feedback in the form of rewards or punishments for its actions. Reinforcement learning has many potential benefits for society, some of which are discussed below. Firstly, reinforcement learning can be used to improve industrial automation and robotics. By training robots to learn from their own experiences, they can gain the skills necessary to perform complex tasks without human intervention. This can lead to increased efficiency and productivity in industries such as manufacturing and logistics. Secondly, reinforcement learning can be used to optimize traffic control systems. By training models to make real-time decisions based on traffic patterns and other data, traffic flow can be improved, reducing congestion and travel times. Thirdly, reinforcement learning can be used to improve healthcare. By training models to make decisions based on patient data, doctors can make more accurate diagnoses and develop more effective treatment plans. This can lead to better health outcomes for patients and can reduce healthcare costs. Fourthly, reinforcement learning can be used to improve education. By training models to adapt to individual student needs, personalized learning experiences can be created that are tailored to each student\u2019s strengths and weaknesses. This can lead to improved academic performance and can help to close the achievement gap. Finally, reinforcement learning can be used to improve environmental sustainability. By training models to make decisions based on environmental data, such as weather patterns and pollution levels, more effective policies can be developed to reduce carbon emissions and protect natural resources. In conclusion, reinforcement learning has many potential benefits for society. By training models to make decisions based on feedback from their environment, we can create more efficient and effective systems in a wide range of fields. However, it is important to consider the ethical implications of these technologies and to ensure that they are developed and used in a responsible and ethical manner. Multi-modal models are another type of machine learning that can process and find relationships between different types of data, such as images, video, audio, and text. They have the potential to revolutionize many aspects of our lives, from healthcare to transportation to education. In this essay, I will discuss how multi-modal models can help society in various ways. One of the most significant benefits of multi-modal models is their ability to transform unstructured data into structured data that can be analyzed. For example, a company could use a multi-modal model to extract data from images or PDFs of invoices or receipts. This would enable them to analyze the data more efficiently and make better-informed decisions. Another benefit of multi-modal models is their ability to cater to various learning styles. Blended and multi-modal learning can reach people who benefit from different learning styles. By understanding their individual learning styles, employees can leverage resources that are compatible with how they process information most effectively. Multi-modal models can also help improve healthcare. For example, they can be used to analyze medical images and identify patterns that might be difficult for human doctors to detect. This can lead to earlier diagnoses and more effective treatments. In addition, multi-modal models can help improve transportation. For example, they can be used to analyze traffic patterns and optimize traffic flow. This can help reduce congestion and improve safety on the roads. Finally, multi-modal models can help improve education. For example, they can be used to create personalized learning experiences for students based on their individual learning styles. This can help students learn more effectively and efficiently. In conclusion, multi-modal models have the potential to help society in many ways. They can transform unstructured data into structured data, cater to various learning styles, improve healthcare, transportation, and education. However, like any new technology, it is important to approach it with caution and consider the potential risks and benefits. I hope this essay has provided some insight into the potential benefits of multi-modal models. Throughout this essay, I have demonstrated the numerous benefits that artificial intelligence will bring to our society. I have also shown some examples of various categories within artificial intelligence that have varying purposes. It is important to consider that each category has its own purpose and has its own pros and cons to it. In conclusion, we must use AI responsibly. "
+}