Share and discover more about AI with social posts from the community.huggingface/OpenAi
Just dropped a fresh version of dataset-viber along with some cool, Gradio-based annotators! These tools aren't about formalities—they're here to help you quickly collect feedback and get your projects moving along to a more serious stage, ahumm @argilla.
Some new features!
- manual import from a CSV or the Hugging Face Hub
- manual export to CSV or the Hub
- improved automated export to the Hub and CSV
Some new features!
- manual import from a CSV or the Hugging Face Hub
- manual export to CSV or the Hub
- improved automated export to the Hub and CSV
When On-Premise is Better than the Cloud
During my time at Palantir, I have spent significant time deploying our software in cloud environments and also a good chunk of time deploying our software in on-premise (on-prem) environments (including starting a team doing just that). I have noticed that despite the common preference for cloud deployment, there are still merits to deploying on-prem.
The Shift from On-Prem to Cloud Computing
Over recent years, the IT landscape has increasingly favored cloud computing, driven by the flexibility of Infrastructure-as-a-Service (IaaS) and Platform-as-a-Service (PaaS) offerings. The global cloud computing market grew from $24.63 billion in 2010 to $156.4 billion in 2020, and that trend continues and is predicted to surpass $1 trillion by 2028. This meteoric rise is powered by both new demand of compute by the world, but also migration of on-prem workflows to the cloud.
There are good reasons for this shift, the cloud enables rapid provisioning of resources, geographic redundancy, and a shift from capital expenditures (CapEx) to operational expenditures (OpEx). However, I believe that there are certain scenarios where the use of on-prem infrastructure, particularly where specific technical requirements, such as deterministic latency, hardware-level control, and stringent security measures, are paramount.
During my time at Palantir, I have spent significant time deploying our software in cloud environments and also a good chunk of time deploying our software in on-premise (on-prem) environments (including starting a team doing just that). I have noticed that despite the common preference for cloud deployment, there are still merits to deploying on-prem.
The Shift from On-Prem to Cloud Computing
Over recent years, the IT landscape has increasingly favored cloud computing, driven by the flexibility of Infrastructure-as-a-Service (IaaS) and Platform-as-a-Service (PaaS) offerings. The global cloud computing market grew from $24.63 billion in 2010 to $156.4 billion in 2020, and that trend continues and is predicted to surpass $1 trillion by 2028. This meteoric rise is powered by both new demand of compute by the world, but also migration of on-prem workflows to the cloud.
There are good reasons for this shift, the cloud enables rapid provisioning of resources, geographic redundancy, and a shift from capital expenditures (CapEx) to operational expenditures (OpEx). However, I believe that there are certain scenarios where the use of on-prem infrastructure, particularly where specific technical requirements, such as deterministic latency, hardware-level control, and stringent security measures, are paramount.
Access and Distribution:
Think of vaccine distribution like a pizza delivery service—except the stakes are much higher, and there are way more “no-delivery zones.” Low-income and remote areas often struggle to get vaccines where they’re needed most. It’s like ordering a pizza to the middle of the Sahara—tricky and sounds impossible, right?
Vaccine Hesitancy:
Ah, the kryptonite of public health. Despite overwhelming evidence, some folks are still hesitant to roll up their sleeves. Whether it’s misinformation on social media or just plain fear of needles, vaccine hesitancy is a real buzzkill. I hope you remember the times when UNICEF used all the tricks held up its sleeves to build confidence in COVID-19 vaccines in the Indian subcontinent.
Think of vaccine distribution like a pizza delivery service—except the stakes are much higher, and there are way more “no-delivery zones.” Low-income and remote areas often struggle to get vaccines where they’re needed most. It’s like ordering a pizza to the middle of the Sahara—tricky and sounds impossible, right?
Vaccine Hesitancy:
Ah, the kryptonite of public health. Despite overwhelming evidence, some folks are still hesitant to roll up their sleeves. Whether it’s misinformation on social media or just plain fear of needles, vaccine hesitancy is a real buzzkill. I hope you remember the times when UNICEF used all the tricks held up its sleeves to build confidence in COVID-19 vaccines in the Indian subcontinent.
A Quick Flashback into Traditional Immunization
Before AI swooped in to save the day, immunization was a bit like the first car our dads got us —effective but sometimes slow and prone to breakdowns. Traditional immunization practices have done wonders, but they’ve also hit a few speed bumps along the way. Remember when getting vaccines to remote areas was about as tricky as finding a parking spot at Coachella? Yeah, that was one of the major challenges. And despite the lifesaving potential, vaccines haven’t always had it easy in the PR department—thanks to the ever-persistent issue of vaccine hesitancy.
Numbers Don’t Lie - Know What’s at Stake
Vaccination rates have improved over the years, but the global picture is still a mixed bag. According to a report by UNICEF, around 23 million children missed out on basic vaccines in 2020—a sharp rise from previous years. While organizations like UNICEF and the CDC are working tirelessly to improve these stats, challenges like distribution, hesitancy, and data management keep throwing wrenches into the works.
But do these challenges continue to be real dealbreakers globally?
Before AI swooped in to save the day, immunization was a bit like the first car our dads got us —effective but sometimes slow and prone to breakdowns. Traditional immunization practices have done wonders, but they’ve also hit a few speed bumps along the way. Remember when getting vaccines to remote areas was about as tricky as finding a parking spot at Coachella? Yeah, that was one of the major challenges. And despite the lifesaving potential, vaccines haven’t always had it easy in the PR department—thanks to the ever-persistent issue of vaccine hesitancy.
Numbers Don’t Lie - Know What’s at Stake
Vaccination rates have improved over the years, but the global picture is still a mixed bag. According to a report by UNICEF, around 23 million children missed out on basic vaccines in 2020—a sharp rise from previous years. While organizations like UNICEF and the CDC are working tirelessly to improve these stats, challenges like distribution, hesitancy, and data management keep throwing wrenches into the works.
But do these challenges continue to be real dealbreakers globally?
Role of AI in Immunization: The Shots that Your Body Needs
Alright, folks, let’s talk about something that’s saved more lives than the Avengers combined—immunization. Vaccines have been the unsung heroes in the battle against some of the world’s nastiest villains (Mpox, please don’t bring back the 2020 phase once again). But in this world of snaps and reels, even our trusty vaccines are getting a digital makeover with AI. It’s more than just modern treatment methods; what if I tell you an AI-driven platform predicted the widespread of COVID-19 before anyone else?
Let me explain it, along with the major role AI plays in the current scenario, so sit tight.
This National Immunization Awareness Month, we will look at how the brainiest sidekick is stepping up to the plate in the world of immunization as it is becoming the Tony Stark of the healthcare world—brilliant, innovative, and just a little bit cooler than everyone else.
Alright, folks, let’s talk about something that’s saved more lives than the Avengers combined—immunization. Vaccines have been the unsung heroes in the battle against some of the world’s nastiest villains (Mpox, please don’t bring back the 2020 phase once again). But in this world of snaps and reels, even our trusty vaccines are getting a digital makeover with AI. It’s more than just modern treatment methods; what if I tell you an AI-driven platform predicted the widespread of COVID-19 before anyone else?
Let me explain it, along with the major role AI plays in the current scenario, so sit tight.
This National Immunization Awareness Month, we will look at how the brainiest sidekick is stepping up to the plate in the world of immunization as it is becoming the Tony Stark of the healthcare world—brilliant, innovative, and just a little bit cooler than everyone else.
Earning potential
According to recent analysis of 342 salaries, a Rust developer in the U.S. makes, on average, $156,000 a year. While the majority of experienced Rustaceans can earn close to $200,000 annually, entry-level positions begin at $121,875 per year – not too shabby.
These figures from job titles including Rust compare well with more general software developer job titles. For example, software engineers command $123,594, system engineers $115,184, and developers $112,502.
Regionally speaking, Texas and New York both offer the highest salaries to Rust developers at $187,500, followed by Georgia ($175,00) and California ($150,000).
Ready to find your next role in tech? Whether you’re a Rust expert or novice or simply want to put your coding expertise to good use, visit the Hackernoon Job Board today
According to recent analysis of 342 salaries, a Rust developer in the U.S. makes, on average, $156,000 a year. While the majority of experienced Rustaceans can earn close to $200,000 annually, entry-level positions begin at $121,875 per year – not too shabby.
These figures from job titles including Rust compare well with more general software developer job titles. For example, software engineers command $123,594, system engineers $115,184, and developers $112,502.
Regionally speaking, Texas and New York both offer the highest salaries to Rust developers at $187,500, followed by Georgia ($175,00) and California ($150,000).
Ready to find your next role in tech? Whether you’re a Rust expert or novice or simply want to put your coding expertise to good use, visit the Hackernoon Job Board today
Companies using Rust
Rust is becoming more and more popular among businesses of all sizes, due to its distinctive qualities, but this is especially true for safety-critical projects. Its wide range of applications includes network programming, web development, and system programming.
In addition, there is a growing need for the system language in the fields of app development, blockchain, Internet of Things, and smart contract programming.
Discord, for instance, accelerates its system by utilizing the low-level language. The chat platform's speed increased tenfold after converting to Rust.
The programming language was used by Meta to make changes to the internal source code management software that its engineers utilize. Dropbox synchronizes files between user devices and its cloud storage via the system language.
Rust is a key part of Microsoft and Amazon’s future, while the U.S. government is even advising to lessen "vulnerabilities at scale," programmers should convert to memory-safe languages like Rust.
Rust is becoming more and more popular among businesses of all sizes, due to its distinctive qualities, but this is especially true for safety-critical projects. Its wide range of applications includes network programming, web development, and system programming.
In addition, there is a growing need for the system language in the fields of app development, blockchain, Internet of Things, and smart contract programming.
Discord, for instance, accelerates its system by utilizing the low-level language. The chat platform's speed increased tenfold after converting to Rust.
The programming language was used by Meta to make changes to the internal source code management software that its engineers utilize. Dropbox synchronizes files between user devices and its cloud storage via the system language.
Rust is a key part of Microsoft and Amazon’s future, while the U.S. government is even advising to lessen "vulnerabilities at scale," programmers should convert to memory-safe languages like Rust.
Rust 411
Rust is expected to be in great demand as a systems language as it is versatile, and used to develop low-level system components as diverse as operating systems, system utilities, device drivers, game and VR simulation engines, and Internet of Things devices.
The language started as a side project for a single Mozilla engineer who intended to create a new programming language that would solve the memory management and allocation issues with C and C++. But later, the open-source software company used Rust as the foundation for a new Firefox browser engine, and a love affair began.
Due to its special qualities, Rust is becoming more and more popular, despite not having the same support ecosystem as older programming languages. However, the systems language has advanced quickly in recent years.
Rust is unique in that it has an ownership and borrowing system that is just as fast and compact as C and C++, but it provides memory security without the need for garbage collection.
In contrast to previous programming languages, Rust guards against memory issues like data races and buffer overflows, and its programmers are protected against mistakes that could result in memory errors by strict data typing constraints. Additionally, its contemporary syntax and overhead-free abstractions have also made a mark.
Rust is expected to be in great demand as a systems language as it is versatile, and used to develop low-level system components as diverse as operating systems, system utilities, device drivers, game and VR simulation engines, and Internet of Things devices.
The language started as a side project for a single Mozilla engineer who intended to create a new programming language that would solve the memory management and allocation issues with C and C++. But later, the open-source software company used Rust as the foundation for a new Firefox browser engine, and a love affair began.
Due to its special qualities, Rust is becoming more and more popular, despite not having the same support ecosystem as older programming languages. However, the systems language has advanced quickly in recent years.
Rust is unique in that it has an ownership and borrowing system that is just as fast and compact as C and C++, but it provides memory security without the need for garbage collection.
In contrast to previous programming languages, Rust guards against memory issues like data races and buffer overflows, and its programmers are protected against mistakes that could result in memory errors by strict data typing constraints. Additionally, its contemporary syntax and overhead-free abstractions have also made a mark.
Mozilla to Meta, Amazon and Microsoft, Rustaceans Are in Demand Right Now
Software developers are feeling the heat, and not just because it’s summertime.
Even though there is currently a high demand for programmers, the rumor mill is turning, and it’s saying that AI may soon replace developers for a sizable chunk of their common tasks.
This may be true of repetitive work and some quality testing, but fortunately, market analysts anticipate that there will be a strong demand for experienced developers in the upcoming years, particularly those who know how to leverage AI.
The future may be uncertain, but one thing we do know is: successful developers will require new skills to be valued by many organizations.
3 high-paying roles to apply for today
ETS Enterprise Portfolio Architect – ISD Prin Arch - Cloud, Navy Federal Credit Union, Winchester ($129,100 - $229,925)
Sr R&D Engineer, Lowe's, Kirkland ($163,800 - $311,200)
Senior DevSecOps Engineer, SciTec, Boulder ($125,000 - $168,400)
Software developers are feeling the heat, and not just because it’s summertime.
Even though there is currently a high demand for programmers, the rumor mill is turning, and it’s saying that AI may soon replace developers for a sizable chunk of their common tasks.
This may be true of repetitive work and some quality testing, but fortunately, market analysts anticipate that there will be a strong demand for experienced developers in the upcoming years, particularly those who know how to leverage AI.
The future may be uncertain, but one thing we do know is: successful developers will require new skills to be valued by many organizations.
3 high-paying roles to apply for today
ETS Enterprise Portfolio Architect – ISD Prin Arch - Cloud, Navy Federal Credit Union, Winchester ($129,100 - $229,925)
Sr R&D Engineer, Lowe's, Kirkland ($163,800 - $311,200)
Senior DevSecOps Engineer, SciTec, Boulder ($125,000 - $168,400)
All the Deadpool and Wolverine Cameos
Happy Hogan
https://marvel-cinematic-universe-guide.fandom.com/wiki/Happy_Hogan
https://marvel-cinematic-universe-guide.fandom.com/wiki/Happy_Hogan
One of the O.G.s of the MCU, Happy Hogan has been a part of the universe since the very first Iron Man. His character has evolved throughout the years, and in Deadpool and Wolverine, he is seen managing the Avengers.
Happy Hogan
https://marvel-cinematic-universe-guide.fandom.com/wiki/Happy_Hogan
https://marvel-cinematic-universe-guide.fandom.com/wiki/Happy_Hogan
One of the O.G.s of the MCU, Happy Hogan has been a part of the universe since the very first Iron Man. His character has evolved throughout the years, and in Deadpool and Wolverine, he is seen managing the Avengers.
Every Deadpool and Wolverine Cameo in Order
Deadpool and Wolverine is currently taking the world by storm. It has officially surpassed the $1 billion mark and has dethroned Joker to become the highest-grossing rated R film of all time. Everyone’s talking about it, and the one thing that people are talking about the most is all the cameos.
MCU movies, especially now that they’re in their Multiverse phase, are filled with a litany of cameos and easter eggs, and Deadpool and Wolverine is no exception. With so many appearing throughout the film, even the most eagle-eyed MCU fan would find it hard to spot them all out.
So, to help you out, here are all the Deadpool and Wolverine Cameos in order.
Want to share your own thoughts on popular media? Start publishing on HackerNoon today!
Deadpool and Wolverine is currently taking the world by storm. It has officially surpassed the $1 billion mark and has dethroned Joker to become the highest-grossing rated R film of all time. Everyone’s talking about it, and the one thing that people are talking about the most is all the cameos.
MCU movies, especially now that they’re in their Multiverse phase, are filled with a litany of cameos and easter eggs, and Deadpool and Wolverine is no exception. With so many appearing throughout the film, even the most eagle-eyed MCU fan would find it hard to spot them all out.
So, to help you out, here are all the Deadpool and Wolverine Cameos in order.
Want to share your own thoughts on popular media? Start publishing on HackerNoon today!
Microsoft releases powerful new Phi-3.5 models, beating Google, OpenAI and more
The three new Phi 3.5 models include the 3.82 billion parameter Phi-3.5-mini-instruct, the 41.9 billion parameter Phi-3.5-MoE-instruct, and the 4.15 billion parameter Phi-3.5-vision-instruct, each designed for basic/fast reasoning, more powerful reasoning, and vision (image and video analysis) tasks, respectively.
All three models are available for developers to download, use, and fine-tune customize on Hugging Face under a Microsoft-branded MIT License that allows for commercial usage and modification without restrictions.
Amazingly, all three models also boast near state-of-the-art performance across a number of third-party benchmark tests, even beating other AI providers including Google’s Gemini 1.5 Flash, Meta’s Llama 3.1, and even OpenAI’s GPT-4o in some cases.
The three new Phi 3.5 models include the 3.82 billion parameter Phi-3.5-mini-instruct, the 41.9 billion parameter Phi-3.5-MoE-instruct, and the 4.15 billion parameter Phi-3.5-vision-instruct, each designed for basic/fast reasoning, more powerful reasoning, and vision (image and video analysis) tasks, respectively.
All three models are available for developers to download, use, and fine-tune customize on Hugging Face under a Microsoft-branded MIT License that allows for commercial usage and modification without restrictions.
Amazingly, all three models also boast near state-of-the-art performance across a number of third-party benchmark tests, even beating other AI providers including Google’s Gemini 1.5 Flash, Meta’s Llama 3.1, and even OpenAI’s GPT-4o in some cases.
huggingface.co
microsoft/Phi-3.5-mini-instruct · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Nvidia’s Llama-3.1-Minitron 4B is a small language model that punches above its weight
As tech companies race to deliver on-device AI, we are seeing a growing body of research and techniques for creating small language models (SLMs) that can run on resource-constrained devices.
The latest models, created by a research team at Nvidia, leverage recent advances in pruning and distillation to create Llama-3.1-Minitron 4B, a compressed version of the Llama 3 model. This model rivals the performance of both larger models and equally sized SLMs while being significantly more efficient to train and deploy.
As tech companies race to deliver on-device AI, we are seeing a growing body of research and techniques for creating small language models (SLMs) that can run on resource-constrained devices.
The latest models, created by a research team at Nvidia, leverage recent advances in pruning and distillation to create Llama-3.1-Minitron 4B, a compressed version of the Llama 3 model. This model rivals the performance of both larger models and equally sized SLMs while being significantly more efficient to train and deploy.
VentureBeat
Why small language models are the next big thing in AI
The specialized nature of SLMs can lead to improved performance and efficiency in these targeted applications compared to using a more general model.
Mozilla/Whisperfile: Local OpenAI Whisper Alternative is Here?Wanna try out FLUX.1 the next generation AI image generator? 🚀🚀🚀
Look no further, Anakin AI offers a whole Universe of AI tools including FLUX.1, DALL.E 3, Stable Diffusion 3, and hundreds of AI tools. So, don’t waste any more time by jumping from website to website.🔥🔥🔥
Experience FLUX, DALLE and Stable Diffusion 3 Now at Anakin AI 👇👇👇
Anakin.ai — One-Stop AI App Platform
Generate Content, Images, Videos, and Voice; Craft Automated Workflows, Custom AI Apps, and Intelligent Agents. Your…
app.anakin.ai
Look no further, Anakin AI offers a whole Universe of AI tools including FLUX.1, DALL.E 3, Stable Diffusion 3, and hundreds of AI tools. So, don’t waste any more time by jumping from website to website.🔥🔥🔥
Experience FLUX, DALLE and Stable Diffusion 3 Now at Anakin AI 👇👇👇
Anakin.ai — One-Stop AI App Platform
Generate Content, Images, Videos, and Voice; Craft Automated Workflows, Custom AI Apps, and Intelligent Agents. Your…
app.anakin.ai
OpenAI GLIDE: Guided Language to Image Diffusion for Generation and Editing
Diffusion Models
Diffusion models work by gradually transforming a noisy image into a clear, detailed one. The process starts with a random noise image, and the model iteratively reduces the noise, guided by the input data, until it produces a realistic image.
Example: Think of it as sculpting a statue from a block of marble. The initial noise represents the unformed block, and each step of the diffusion process chisels away the noise, revealing the final image.
Diffusion Models
Diffusion models work by gradually transforming a noisy image into a clear, detailed one. The process starts with a random noise image, and the model iteratively reduces the noise, guided by the input data, until it produces a realistic image.
Example: Think of it as sculpting a statue from a block of marble. The initial noise represents the unformed block, and each step of the diffusion process chisels away the noise, revealing the final image.
Doc to Dialogue in Hugging Face
Project
Transform any PDF document (research report, market analysis, manuals, or user guides) into an audio interview with two AI-generated voices to enhance engagement with complex content. I used the Gemini API model for document processing, OpenAI Whisper TTS for voice generation, and Gradio for the interface, and uploaded in huggingface.
Any feedback will be welcome!
Project
Transform any PDF document (research report, market analysis, manuals, or user guides) into an audio interview with two AI-generated voices to enhance engagement with complex content. I used the Gemini API model for document processing, OpenAI Whisper TTS for voice generation, and Gradio for the interface, and uploaded in huggingface.
Any feedback will be welcome!
State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow
🤗 Transformers provides thousands of pretrained models to perform tasks on different modalities such as text, vision, and audio.
These models can be applied on:
• 📝 Text, for tasks like text classification, information extraction, question answering, summarization, translation, and text generation, in over 100 languages.
• 🖼️ Images, for tasks like image classification, object detection, and segmentation.
• 🗣️ Audio, for tasks like speech recognition and audio classification.
Transformer models can also perform tasks on several modalities combined, such as table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
🤗 Transformers provides APIs to quickly download and use those pretrained models on a given text, fine-tune them on your own datasets and then share them with the community on our model hub. At the same time, each python module defining an architecture is fully standalone and can be modified to enable quick research experiments.
🤗 Transformers provides thousands of pretrained models to perform tasks on different modalities such as text, vision, and audio.
These models can be applied on:
• 📝 Text, for tasks like text classification, information extraction, question answering, summarization, translation, and text generation, in over 100 languages.
• 🖼️ Images, for tasks like image classification, object detection, and segmentation.
• 🗣️ Audio, for tasks like speech recognition and audio classification.
Transformer models can also perform tasks on several modalities combined, such as table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
🤗 Transformers provides APIs to quickly download and use those pretrained models on a given text, fine-tune them on your own datasets and then share them with the community on our model hub. At the same time, each python module defining an architecture is fully standalone and can be modified to enable quick research experiments.
huggingface.co
Models - Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Everything you need to know about using the tools, libraries, and models at Hugging Face—from transformers, to RAG, LangChain, and Gradio.
Hugging Face is the ultimate resource for machine learning engineers and AI developers. It provides hundreds of pretrained and open-source models for dozens of different domains—from natural language processing to computer vision. Plus, you’ll find a popular platform for hosting your models and datasets. Hugging Face in Action reveals how to get the absolute best out of everything Hugging Face, from accessing state-of-the-art models to building intuitive frontends for AI apps.
Hugging Face is the ultimate resource for machine learning engineers and AI developers. It provides hundreds of pretrained and open-source models for dozens of different domains—from natural language processing to computer vision. Plus, you’ll find a popular platform for hosting your models and datasets. Hugging Face in Action reveals how to get the absolute best out of everything Hugging Face, from accessing state-of-the-art models to building intuitive frontends for AI apps.
microsoft/Phi-3.5-MoE-instruct
Model Summary
Phi-3.5-MoE is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available documents - with a focus on very high-quality, reasoning dense data. The model supports multilingual and comes with 128K context length (in tokens). The model underwent a rigorous enhancement process, incorporating supervised fine-tuning, proximal policy optimization, and direct preference optimization to ensure precise instruction adherence and robust safety measures.
🏡 Phi-3 Portal 📰 Phi-3 Microsoft Blog 📖 Phi-3 Technical Report 👩🍳 Phi-3 Cookbook 🖥️ Try It
Model Summary
Phi-3.5-MoE is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available documents - with a focus on very high-quality, reasoning dense data. The model supports multilingual and comes with 128K context length (in tokens). The model underwent a rigorous enhancement process, incorporating supervised fine-tuning, proximal policy optimization, and direct preference optimization to ensure precise instruction adherence and robust safety measures.
🏡 Phi-3 Portal 📰 Phi-3 Microsoft Blog 📖 Phi-3 Technical Report 👩🍳 Phi-3 Cookbook 🖥️ Try It
microsoft/Phi-3.5-vision-instruct Model Summary
Phi-3.5-vision is a lightweight, state-of-the-art open multimodal model built upon datasets which include - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data both on text and vision. The model belongs to the Phi-3 model family, and the multimodal version comes with 128K context length (in tokens) it can support. The model underwent a rigorous enhancement process, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures.
🏡 Phi-3 Portal 📰 Phi-3 Microsoft Blog 📖 Phi-3 Technical Report 👩🍳 Phi-3 Cookbook 🖥️ Try It
Phi-3.5-vision is a lightweight, state-of-the-art open multimodal model built upon datasets which include - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data both on text and vision. The model belongs to the Phi-3 model family, and the multimodal version comes with 128K context length (in tokens) it can support. The model underwent a rigorous enhancement process, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures.
🏡 Phi-3 Portal 📰 Phi-3 Microsoft Blog 📖 Phi-3 Technical Report 👩🍳 Phi-3 Cookbook 🖥️ Try It