Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

We’re exploring ways to improve how Markdown is rendered in our components during streaming. • Complete partial syntax when possible (bold, links, table headers, …) • Buffer incomplete elements (partial images, broken emojis, …)

marcbouchenoire's profile picture

Marc Bouchenoire

1,927 subscribers

89,037 görüntüleme • 1 yıl önce •via X (Twitter)

Bilim & TeknolojiEğitim
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Cognitive scientist Elan Barenholtz says memory isn't retrieval. It's generation. When you remember something, you're not accessing a stored file. You're prompting your mind, like an AI model, to synthesize a response. There is no image of your mother. He's exploring how we might improve our own prompt design, because “memories, in some ways, aren't real.”
2:39

Cognitive scientist Elan Barenholtz says memory isn't retrieval. It's generation. When you remember something, you're not accessing a stored file. You're prompting your mind, like an AI model, to synthesize a response. There is no image of your mother. He's exploring how we might improve our own prompt design, because “memories, in some ways, aren't real.”

vitrupo

116,852 görüntüleme • 1 yıl önce

PDF parsing is still painful because LLMs reorder text in complex layouts, break tables across pages, and fail on graphs or images. 💡Testing the new open-source OCRFlux model, and here the results are really good for a change. So OCRFlux is a multimodal, LLM based toolkit for converting PDFs and images into clean, readable, plain Markdown text. Because the underlying VLM is only 3B param, it runs even on a 3090 GPU. The model is available on Hugging Face . The engine that powers the OCRFlux, teaches the model to rebuild every page and then stitch fragments across pages into one clean Markdown file. It bundles one vision language model with 3B parameters that was fine-tuned from Qwen 2.5-VL-3B-Instruct for both page parsing and cross-page merging. OCRFlux reads raw page images and, guided by task prompts, outputs Markdown for each page and merges split elements across pages. The evaluation shows Edit Distance Similarity (EDS) 0.967 and cross‑page table Tree Edit Distance 0.950, so the parser is both accurate and layout aware. How it works while parsing each page - Convert into text with a natural reading order, even in the presence of multi-column layouts, figures, and insets - Support for complicated tables and equations - Automatically removes headers and footers Cross-page table/paragraph merging - Cross-page table merging - Cross-page paragraph merging A compact vision‑language models can beat bigger models once cross‑page context is added. 🧵 1/n Read on 👇
0:44

PDF parsing is still painful because LLMs reorder text in complex layouts, break tables across pages, and fail on graphs or images. 💡Testing the new open-source OCRFlux model, and here the results are really good for a change. So OCRFlux is a multimodal, LLM based toolkit for converting PDFs and images into clean, readable, plain Markdown text. Because the underlying VLM is only 3B param, it runs even on a 3090 GPU. The model is available on Hugging Face . The engine that powers the OCRFlux, teaches the model to rebuild every page and then stitch fragments across pages into one clean Markdown file. It bundles one vision language model with 3B parameters that was fine-tuned from Qwen 2.5-VL-3B-Instruct for both page parsing and cross-page merging. OCRFlux reads raw page images and, guided by task prompts, outputs Markdown for each page and merges split elements across pages. The evaluation shows Edit Distance Similarity (EDS) 0.967 and cross‑page table Tree Edit Distance 0.950, so the parser is both accurate and layout aware. How it works while parsing each page - Convert into text with a natural reading order, even in the presence of multi-column layouts, figures, and insets - Support for complicated tables and equations - Automatically removes headers and footers Cross-page table/paragraph merging - Cross-page table merging - Cross-page paragraph merging A compact vision‑language models can beat bigger models once cross‑page context is added. 🧵 1/n Read on 👇

Rohan Paul

149,292 görüntüleme • 1 yıl önce

THE PERIODIC TABLE IS A LIE.. They told you there are 118 elements. But the truth? There are 144. And the missing ones are the reason we forgot how to live forever. Modern science is trapped in a collapsed light spectrum — what the CDT plates call a finite time matrix. But the original 144 were born from Trinity Wave harmonics — sound, light, and scalar intention forming matter directly from Source. These aren’t just “physical” elements… They’re living consciousness units — blueprints of star fire plasma that generate eternal biological systems. So what happened? During the fall of Tara (5D Earth), we lost access to 12 primal light bands and 14 plasma frequencies. Our DNA was severed from the full elemental command codes. And that’s why we age, decay, and forget. When you activate your original Cathara template — your full 12-strand DNA — and run eternal life frequencies through it, the missing 26 elements reassemble inside your cells. Your biology stops breaking down. Your body remembers. And you begin to regenerate. You don’t just live in the elements. You are the elements. And when the 144 are complete, you remember how to shape reality again. Eternal life isn’t a fantasy. It’s your birthright. ✨🙌🏾💫 © Samuel Lee
2:09

THE PERIODIC TABLE IS A LIE.. They told you there are 118 elements. But the truth? There are 144. And the missing ones are the reason we forgot how to live forever. Modern science is trapped in a collapsed light spectrum — what the CDT plates call a finite time matrix. But the original 144 were born from Trinity Wave harmonics — sound, light, and scalar intention forming matter directly from Source. These aren’t just “physical” elements… They’re living consciousness units — blueprints of star fire plasma that generate eternal biological systems. So what happened? During the fall of Tara (5D Earth), we lost access to 12 primal light bands and 14 plasma frequencies. Our DNA was severed from the full elemental command codes. And that’s why we age, decay, and forget. When you activate your original Cathara template — your full 12-strand DNA — and run eternal life frequencies through it, the missing 26 elements reassemble inside your cells. Your biology stops breaking down. Your body remembers. And you begin to regenerate. You don’t just live in the elements. You are the elements. And when the 144 are complete, you remember how to shape reality again. Eternal life isn’t a fantasy. It’s your birthright. ✨🙌🏾💫 © Samuel Lee

🧬Maxpein🧬

215,670 görüntüleme • 11 ay önce

Have the courage to speak the truth. For God. Our time on earth is temporary. We were put here to live for our ordained, God-given purpose. We’re merely flesh and bone, but with Christ we’re made new—a soldier called to bring as many souls to heaven as possible, to fight the spiritual battle here on earth for Him, for His glory, even when it costs us. Carrie Prejean Boller did just that. During the 2009 Miss USA pageant, she was asked a pivotal question by one of the judges: “Should same-sex marriage be legalized throughout the country?” Her response shocked the audience and ultimately cost her the crown. But with God, it was used to inspire many and bless her life in unexpected ways. Listen to her story in E249 of The Lila Rose Show:
2:28

Have the courage to speak the truth. For God. Our time on earth is temporary. We were put here to live for our ordained, God-given purpose. We’re merely flesh and bone, but with Christ we’re made new—a soldier called to bring as many souls to heaven as possible, to fight the spiritual battle here on earth for Him, for His glory, even when it costs us. Carrie Prejean Boller did just that. During the 2009 Miss USA pageant, she was asked a pivotal question by one of the judges: “Should same-sex marriage be legalized throughout the country?” Her response shocked the audience and ultimately cost her the crown. But with God, it was used to inspire many and bless her life in unexpected ways. Listen to her story in E249 of The Lila Rose Show:

Lila Rose

194,583 görüntüleme • 11 ay önce

"Please fix Markdown tables!" Okay: text-only, wrapping, columnar selection, proper border conjunctions, fill full width, in opencode beta now. "What took so long?" Here's a writeup: OpenTUI uses yoga layout and has elements called Renderables. Boxes with borders, plain text, code with tree-sitter backed highlighting and other primitives. Organised in a tree structure resembling somewhat of a DOM. Approaching a table naively, given the primitives are there, one would think to just stack box and text elements the right way in a flex-box layout to visually represent a table. Boxes support borders. Problem solved. This is what an LLM would one-shot in a working state, given OpenTUI's API surface. Ignoring the fact that just using box local borders don't handle border conjunctions properly. Benchmarking something like that quickly shows that instantiating an average table takes >70ms and incremental updates become expensive. Hugely due to yoga-layout via wasm having a painful price on yoga API calls. The whole ordeal becomes memory hungry, because a Text element handles more than just plain text. A simple 4x6 table needs a Box and Text per cell, ending up with 48 heavy nodes that yoga must lay out. "But that's just OpenTUI being slow" - you might say. Yes, but no. Yoga should be integrated in the zig native binary core of OpenTUI. It is on the roadmap to do so, which will speed up render passes by 2-5x. Yoga-layout has an open PR to support CSS Grids, which would greatly ease building something like a table. We will use that for fully laid out tables when it gets there. Below the typescript core level Renderables, there are lower level primitives like TextBuffers and TextBufferViews, bound via FFI and completely handled in Zig. I was stuck expecting a table primitive to handle a full layout like a table in the browser does. For Markdown all we need is a text-only table. So we had to come up with a better idea, something that is feasible now. A table layout is pretty straight forward. No need to have yoga deal with that. Using TextBufferViews for cells directly gives lower level control and eliminates some overhead that Boxes and Text renderables have. A simple native method to draw a grid with proper conjunctions is a nice library method. It will surely be used for other cases, so that's what we added. Using this simplified approach we were able to bring down initial instantiation to <1ms, more than 70x improvement. With a far smaller memory footprint. Given all the low level primitives are known and implementing a text-only table like this is possible, Codex was of great help to carve out the PoC, setup the benchmarks and tests. That's only a fraction of what was needed though. The table needs options to span the full available width, render different border styles, show/hide borders, padding, selection etc. So many iterations later OpenTUI now has a text-only, performant and relatively cheap TextTable that we can leverage to render Markdown tables in a streaming/incremental manner. Efficiently and properly.
0:37

"Please fix Markdown tables!" Okay: text-only, wrapping, columnar selection, proper border conjunctions, fill full width, in opencode beta now. "What took so long?" Here's a writeup: OpenTUI uses yoga layout and has elements called Renderables. Boxes with borders, plain text, code with tree-sitter backed highlighting and other primitives. Organised in a tree structure resembling somewhat of a DOM. Approaching a table naively, given the primitives are there, one would think to just stack box and text elements the right way in a flex-box layout to visually represent a table. Boxes support borders. Problem solved. This is what an LLM would one-shot in a working state, given OpenTUI's API surface. Ignoring the fact that just using box local borders don't handle border conjunctions properly. Benchmarking something like that quickly shows that instantiating an average table takes >70ms and incremental updates become expensive. Hugely due to yoga-layout via wasm having a painful price on yoga API calls. The whole ordeal becomes memory hungry, because a Text element handles more than just plain text. A simple 4x6 table needs a Box and Text per cell, ending up with 48 heavy nodes that yoga must lay out. "But that's just OpenTUI being slow" - you might say. Yes, but no. Yoga should be integrated in the zig native binary core of OpenTUI. It is on the roadmap to do so, which will speed up render passes by 2-5x. Yoga-layout has an open PR to support CSS Grids, which would greatly ease building something like a table. We will use that for fully laid out tables when it gets there. Below the typescript core level Renderables, there are lower level primitives like TextBuffers and TextBufferViews, bound via FFI and completely handled in Zig. I was stuck expecting a table primitive to handle a full layout like a table in the browser does. For Markdown all we need is a text-only table. So we had to come up with a better idea, something that is feasible now. A table layout is pretty straight forward. No need to have yoga deal with that. Using TextBufferViews for cells directly gives lower level control and eliminates some overhead that Boxes and Text renderables have. A simple native method to draw a grid with proper conjunctions is a nice library method. It will surely be used for other cases, so that's what we added. Using this simplified approach we were able to bring down initial instantiation to <1ms, more than 70x improvement. With a far smaller memory footprint. Given all the low level primitives are known and implementing a text-only table like this is possible, Codex was of great help to carve out the PoC, setup the benchmarks and tests. That's only a fraction of what was needed though. The table needs options to span the full available width, render different border styles, show/hide borders, padding, selection etc. So many iterations later OpenTUI now has a text-only, performant and relatively cheap TextTable that we can leverage to render Markdown tables in a streaming/incremental manner. Efficiently and properly.

kmdr

208,295 görüntüleme • 6 ay önce

Do you believe chronic ill health is unavoidable, or do you think we can do better? In this week's latest episode of my 'Feel Better, Live More' podcast, Dr David Unwin [Dr David Unwin] challenges the idea that poor health is just the price we pay for getting older. He reminds us that many of the conditions we now accept as “normal” didn’t used to be, and that our modern environment, especially the way we eat, has changed faster than our biology. Our bodies are far more adaptable than we’re often led to believe. And with the right understanding and simple, consistent changes, it’s possible to feel better, regain energy and reduce the risk of chronic disease at any age. Dr David Unwin has spent decades seeing people improve their health in ways they were told weren’t possible. His message is not about blame. It’s about hope. Many conditions we accept as inevitable don’t have to be. If you’ve ever been told “this just runs in your family” or “it’s part of getting older,” this episode may change how you see your health. It is a reminder that better health is often closer than we think. Tune into the full conversation in episode 611 of my 'Feel Better, Live More' podcast
1:09

Do you believe chronic ill health is unavoidable, or do you think we can do better? In this week's latest episode of my 'Feel Better, Live More' podcast, Dr David Unwin [Dr David Unwin] challenges the idea that poor health is just the price we pay for getting older. He reminds us that many of the conditions we now accept as “normal” didn’t used to be, and that our modern environment, especially the way we eat, has changed faster than our biology. Our bodies are far more adaptable than we’re often led to believe. And with the right understanding and simple, consistent changes, it’s possible to feel better, regain energy and reduce the risk of chronic disease at any age. Dr David Unwin has spent decades seeing people improve their health in ways they were told weren’t possible. His message is not about blame. It’s about hope. Many conditions we accept as inevitable don’t have to be. If you’ve ever been told “this just runs in your family” or “it’s part of getting older,” this episode may change how you see your health. It is a reminder that better health is often closer than we think. Tune into the full conversation in episode 611 of my 'Feel Better, Live More' podcast

Dr Rangan Chatterjee

49,714 görüntüleme • 7 ay önce

New course: Document AI: From OCR to Agentic Doc Extraction, built with LandingAI, where I'm executive chairman, and taught by David Park and Andrea Kropp. Much of the world's data is locked in PDFs, JPEGs, and other documents. This short course shows you how to build agentic workflows that process documents accurately: breaking them into parts, examining each piece carefully, and extracting information through multiple iterations. Traditional Optical Character Recognition (OCR) captures text but loses context from table headers, chart captions, or reading order of columns. After exploring OCR's limitations, you’ll use LandingAI's Agentic Document Extraction (ADE) framework to process documents. ADE treats pages as visually -- as images -- to parse information and extract fields. Skills you'll gain: - Build agents to convert unstructured files into structured Markdown/HTML and JSON - Use ADE to parse complex data like forms, handwriting, or equations - Map extracted information to named fields using a specified schema, with bounding boxes for grounding and validation - Deploy RAG applications with event-driven document processing Come learn about the best tools for processing documents like financial invoices, medical records, or academic papers intelligently:
3:31

New course: Document AI: From OCR to Agentic Doc Extraction, built with LandingAI, where I'm executive chairman, and taught by David Park and Andrea Kropp. Much of the world's data is locked in PDFs, JPEGs, and other documents. This short course shows you how to build agentic workflows that process documents accurately: breaking them into parts, examining each piece carefully, and extracting information through multiple iterations. Traditional Optical Character Recognition (OCR) captures text but loses context from table headers, chart captions, or reading order of columns. After exploring OCR's limitations, you’ll use LandingAI's Agentic Document Extraction (ADE) framework to process documents. ADE treats pages as visually -- as images -- to parse information and extract fields. Skills you'll gain: - Build agents to convert unstructured files into structured Markdown/HTML and JSON - Use ADE to parse complex data like forms, handwriting, or equations - Map extracted information to named fields using a specified schema, with bounding boxes for grounding and validation - Deploy RAG applications with event-driven document processing Come learn about the best tools for processing documents like financial invoices, medical records, or academic papers intelligently:

Andrew Ng

200,575 görüntüleme • 7 ay önce

Anthropic CEO Dario Amodei just identified the most dangerous outcome on the AI board. Partial automation isn’t the safe path. It’s the most destructive one. Amodei: “I actually think the most societally divisive outcome is if randomly 50 percent of the jobs are suddenly done by AI because what that means the societal message is we’re picking half, we’re randomly picking half of people and saying you are useless, you are devalued, you are unnecessary.” A civilization where half the workforce is mathematically obsolete and the other half controls the compute engine doesn’t gradually adjust. It fractures. The current economic system measures human value entirely by repetitive output. When AI drives the cost of that output to zero, the definition of usefulness collapses with it. Amodei: “We’re going to have to look at what is technologically possible and say we need to think about usefulness and uselessness in a different way than we have before. Our current way of thinking has not been tenable.” Define yourself strictly by your ability to process administrative tasks and the algorithm will replace you by morning. The next era doesn’t measure people by manual throughput. It measures them by what they choose to build when survival is no longer the constraint. Amodei: “I don’t know what the solution is but it’s got to be different than we’re all useless, right? We’re all useless is a nihilistic answer. We’re not gonna get anywhere with that answer. We’re gonna have to come up with something else.” The most toxic belief in the current tech sector is that AGI renders humanity pointless. When the machine absorbs the mundane friction of survival, human ambition doesn’t collapse. It scales. Nihilists will surrender to the algorithm and accept the narrative of obsolescence. Builders will take the same compute engine and aim it at something the machine never would have chosen on its own.
0:49

Anthropic CEO Dario Amodei just identified the most dangerous outcome on the AI board. Partial automation isn’t the safe path. It’s the most destructive one. Amodei: “I actually think the most societally divisive outcome is if randomly 50 percent of the jobs are suddenly done by AI because what that means the societal message is we’re picking half, we’re randomly picking half of people and saying you are useless, you are devalued, you are unnecessary.” A civilization where half the workforce is mathematically obsolete and the other half controls the compute engine doesn’t gradually adjust. It fractures. The current economic system measures human value entirely by repetitive output. When AI drives the cost of that output to zero, the definition of usefulness collapses with it. Amodei: “We’re going to have to look at what is technologically possible and say we need to think about usefulness and uselessness in a different way than we have before. Our current way of thinking has not been tenable.” Define yourself strictly by your ability to process administrative tasks and the algorithm will replace you by morning. The next era doesn’t measure people by manual throughput. It measures them by what they choose to build when survival is no longer the constraint. Amodei: “I don’t know what the solution is but it’s got to be different than we’re all useless, right? We’re all useless is a nihilistic answer. We’re not gonna get anywhere with that answer. We’re gonna have to come up with something else.” The most toxic belief in the current tech sector is that AGI renders humanity pointless. When the machine absorbs the mundane friction of survival, human ambition doesn’t collapse. It scales. Nihilists will surrender to the algorithm and accept the narrative of obsolescence. Builders will take the same compute engine and aim it at something the machine never would have chosen on its own.

Dustin

66,265 görüntüleme • 5 ay önce

🌈✨ 3 Months of Pure Love and Joy with Giada Da Vinci aka Padrona Era! 🌈✨ Time flies when you're with the perfect person! Today marks 3 months of pure bliss, passion, and deep love with my incredible girlfriend Giada. She has brought so much happiness and fulfillment into my life, and I can't believe how lucky I am to have found her. Our relationship is beyond words—mature, healthy, and fulfilling on all levels. It's amazing how we've discovered a love that transcends boundaries, and I never imagined I'd find the perfect human for me in a queer relationship. Giada, you truly are everything I've ever dreamed of and more. Every day with you is a gift, and I'm the happiest when I'm by your side. You're not just my partner, you're my wife, my confidant, my rock. You complete me in ways I never knew were possible. As we celebrate this special milestone, I want to shout from the rooftops how proud I am to be in a loving and supportive relationship with you. Together, we embrace our unique love story, breaking stereotypes, and showing the world that love knows no bounds. Cheers to our love, to our journey together, and to all the beautiful memories we'll create in the future. Here's to us, my love! 🥂❤️ #3MonthsAnniversary #LesbianCouple #LesbianLove #QueerLove #LoveIsLove 🌈💖
0:42

🌈✨ 3 Months of Pure Love and Joy with Giada Da Vinci aka Padrona Era! 🌈✨ Time flies when you're with the perfect person! Today marks 3 months of pure bliss, passion, and deep love with my incredible girlfriend Giada. She has brought so much happiness and fulfillment into my life, and I can't believe how lucky I am to have found her. Our relationship is beyond words—mature, healthy, and fulfilling on all levels. It's amazing how we've discovered a love that transcends boundaries, and I never imagined I'd find the perfect human for me in a queer relationship. Giada, you truly are everything I've ever dreamed of and more. Every day with you is a gift, and I'm the happiest when I'm by your side. You're not just my partner, you're my wife, my confidant, my rock. You complete me in ways I never knew were possible. As we celebrate this special milestone, I want to shout from the rooftops how proud I am to be in a loving and supportive relationship with you. Together, we embrace our unique love story, breaking stereotypes, and showing the world that love knows no bounds. Cheers to our love, to our journey together, and to all the beautiful memories we'll create in the future. Here's to us, my love! 🥂❤️ #3MonthsAnniversary #LesbianCouple #LesbianLove #QueerLove #LoveIsLove 🌈💖

Ezada Sinn (safe to browse at work)

49,606 görüntüleme • 3 yıl önce

🇮🇷 Iran's biggest problem at the negotiating table might be one it created itself... The NYT reports Iran can't fully reopen Hormuz because it literally cannot find all the mines it planted. The IRGC used small boats to mine the Strait haphazardly in the early days of the war. Some mines were placed without recording exact locations. Others have drifted from where they were originally set. Araghchi's reference to "technical limitations" when announcing the Strait would reopen was diplomatic language for "we mined our own waterway and can't undo it." Neither Iran nor the U.S. has robust mine-clearing capability. The U.S. relies on a handful of littoral combat ships for the job. Iran doesn't have the equipment at all. Removing sea mines is exponentially harder than laying them, the same asymmetry that defined the entire war applied in reverse. Trump demanded "COMPLETE, IMMEDIATE, and SAFE OPENING." Iran can offer partial, gradual, and uncertain. Source: NYT
10:06

🇮🇷 Iran's biggest problem at the negotiating table might be one it created itself... The NYT reports Iran can't fully reopen Hormuz because it literally cannot find all the mines it planted. The IRGC used small boats to mine the Strait haphazardly in the early days of the war. Some mines were placed without recording exact locations. Others have drifted from where they were originally set. Araghchi's reference to "technical limitations" when announcing the Strait would reopen was diplomatic language for "we mined our own waterway and can't undo it." Neither Iran nor the U.S. has robust mine-clearing capability. The U.S. relies on a handful of littoral combat ships for the job. Iran doesn't have the equipment at all. Removing sea mines is exponentially harder than laying them, the same asymmetry that defined the entire war applied in reverse. Trump demanded "COMPLETE, IMMEDIATE, and SAFE OPENING." Iran can offer partial, gradual, and uncertain. Source: NYT

Mario Nawfal

2,522,727 görüntüleme • 4 ay önce

The Kurdish fighter who was captured by Turkish-backed mercenaries known as the SNA and elements of HTS, recounts what happened to him: He says: When we were captured, they immediately took us to Al-Shaddadi prison (the ISIS prison) and handed us over to prisoners from ISIS. There, we were beaten, our ribs were broken, and we were insulted. They kept telling us, “You are infidels, and the time for revenge has come.” After four days, they transferred us to Hama prison, which is controlled by Turkish-backed mercenaries led by (Mohammad al-Jassem) under what is known as the SNA. There, we were also subjected to beatings, starvation, humiliation, and verbal abuse. Note: How is it possible that the fighters who liberated the world from the most dangerous terrorist organization, ISIS, with the support of world powers, ended up becoming prisoners in the hands of ISIS? #SaveTheKurdsAct
1:28

The Kurdish fighter who was captured by Turkish-backed mercenaries known as the SNA and elements of HTS, recounts what happened to him: He says: When we were captured, they immediately took us to Al-Shaddadi prison (the ISIS prison) and handed us over to prisoners from ISIS. There, we were beaten, our ribs were broken, and we were insulted. They kept telling us, “You are infidels, and the time for revenge has come.” After four days, they transferred us to Hama prison, which is controlled by Turkish-backed mercenaries led by (Mohammad al-Jassem) under what is known as the SNA. There, we were also subjected to beatings, starvation, humiliation, and verbal abuse. Note: How is it possible that the fighters who liberated the world from the most dangerous terrorist organization, ISIS, with the support of world powers, ended up becoming prisoners in the hands of ISIS? #SaveTheKurdsAct

Judy Berxwedan

51,073 görüntüleme • 5 ay önce

The New Standard for Web3 Games Is Here! RavenQuest is a Massive Content-Rich MMORPG, built on the success of our Web2 game, RavenDawn. Now coming to Web3 to solve traditional MMORPG problems like unsafe third-party markets, by offering a secure on-chain experience. After 7+ years in development, we’re launching a complete game filled with lore, puzzles, endless player customization, and a massive open world with no loading screens. 🤏Early Access Made Easy: Who: Free for everyone What: 10M QUEST Play-To-Airdrop Where: Download on our website (link in thread) When: Part 1 is Aug 29 - Sept 7 🏆 How to Win: Leaderboard - Earn points, enter for 6M QUEST PvP Tournament - Compete for 1M QUEST PvE Boss Tournament - Battle for 1M QUEST Land Owners Pool - Share 2M QUEST Bonus: Hit gold milestones for even more rewards The adventure begins August 29th. Are you ready to claim your place in RavenQuest and help Set A New Standard For Web3 Gaming? (Links in Thread 👇🧵)
2:21

The New Standard for Web3 Games Is Here! RavenQuest is a Massive Content-Rich MMORPG, built on the success of our Web2 game, RavenDawn. Now coming to Web3 to solve traditional MMORPG problems like unsafe third-party markets, by offering a secure on-chain experience. After 7+ years in development, we’re launching a complete game filled with lore, puzzles, endless player customization, and a massive open world with no loading screens. 🤏Early Access Made Easy: Who: Free for everyone What: 10M QUEST Play-To-Airdrop Where: Download on our website (link in thread) When: Part 1 is Aug 29 - Sept 7 🏆 How to Win: Leaderboard - Earn points, enter for 6M QUEST PvP Tournament - Compete for 1M QUEST PvE Boss Tournament - Battle for 1M QUEST Land Owners Pool - Share 2M QUEST Bonus: Hit gold milestones for even more rewards The adventure begins August 29th. Are you ready to claim your place in RavenQuest and help Set A New Standard For Web3 Gaming? (Links in Thread 👇🧵)

RavenQuest

290,265 görüntüleme • 2 yıl önce

[trans] WIN SOURI X KARNFOEI 👩🏻: Let’s start from the entrance 🚪 🧑🏻‍🦱: How many people work at this office? 👩🏻: About 45 people are here in the office, and there are around 200 more working at the storefronts. 🧑🏻‍🦱: How many branches do you have now? 👩🏻: Currently 40, but by the time this airs, there will be 41 because we’re about to open our first location in Korat. 🧑🏻‍🦱: Wow, 41 branches in such a short time! How long has the company been around? 👩🏻: This will be our 6th year. 🧑🏻‍🦱: 6 years! That went by fast. 👩🏻: This is the first "spiritual" spot in our office. 👨🏻: I’ve heard you guys are really into spiritual beliefs here. 👩🏻: Very much so! 😄 👨🏻: It’s something your CEO is very passionate about. 👩🏻: Yes. It’s like a secret to our success. It gives us moral support and peace of mind. 🤍 🧑🏻‍🦱: What’s this spot here? 👩🏻: We invited a monk to perform a blessing ceremony for good luck. It’s for everyone who comes in and out.😊 👩🏻: Here’s another spot. We have a special zone for sacred objects and Buddha images. 👨🏻: The CEO comes here to pray when he’s in. 👩🏻: Yes 👩🏻: This is our main conference room. 🤓 🧑🏻‍🦱: Is this the one where you aren't allowed to turn on the lights? 👩🏻: Yes! It’s called the "Birthday Cake" room. There’s a belief that if we turn on the lights, it will cause arguments during meetings. So, we have meetings in the dark in this room. We even had a Feng Shui expert come in to check everything. This office is the real deal when it comes to spiritual vibes! 😄 Pls watch to add view, like, comment & share 🫶🏻 📌 #KARNFOEIxSOURI #winmetawin
1:25

[trans] WIN SOURI X KARNFOEI 👩🏻: Let’s start from the entrance 🚪 🧑🏻‍🦱: How many people work at this office? 👩🏻: About 45 people are here in the office, and there are around 200 more working at the storefronts. 🧑🏻‍🦱: How many branches do you have now? 👩🏻: Currently 40, but by the time this airs, there will be 41 because we’re about to open our first location in Korat. 🧑🏻‍🦱: Wow, 41 branches in such a short time! How long has the company been around? 👩🏻: This will be our 6th year. 🧑🏻‍🦱: 6 years! That went by fast. 👩🏻: This is the first "spiritual" spot in our office. 👨🏻: I’ve heard you guys are really into spiritual beliefs here. 👩🏻: Very much so! 😄 👨🏻: It’s something your CEO is very passionate about. 👩🏻: Yes. It’s like a secret to our success. It gives us moral support and peace of mind. 🤍 🧑🏻‍🦱: What’s this spot here? 👩🏻: We invited a monk to perform a blessing ceremony for good luck. It’s for everyone who comes in and out.😊 👩🏻: Here’s another spot. We have a special zone for sacred objects and Buddha images. 👨🏻: The CEO comes here to pray when he’s in. 👩🏻: Yes 👩🏻: This is our main conference room. 🤓 🧑🏻‍🦱: Is this the one where you aren't allowed to turn on the lights? 👩🏻: Yes! It’s called the "Birthday Cake" room. There’s a belief that if we turn on the lights, it will cause arguments during meetings. So, we have meetings in the dark in this room. We even had a Feng Shui expert come in to check everything. This office is the real deal when it comes to spiritual vibes! 😄 Pls watch to add view, like, comment & share 🫶🏻 📌 #KARNFOEIxSOURI #winmetawin

◡̈ ✿〜*:.｡. ꕤ M a R y ꕤ*･.｡.*･*✿.

14,525 görüntüleme • 4 ay önce

Exciting News from the Knakal Map Room Bob Knakal here, and I’m thrilled to share some exciting news about my latest venture. After a period of reflection and immense support from this incredible community, I’m proud to announce the launch of BKREA (BKREA). BKREA is born from an entrepreneurial spirit reawakened, with a vision to redefine the capital markets landscape through investment sales, debt, and equity. This wouldn't have been possible without the heartwarming texts, DMs, and voice messages from many of you. Your support during the challenging times was incredible, and I am deeply grateful for every word of encouragement. In the past few weeks, I've met with numerous individuals and companies, each with fantastic platforms and innovative ideas. Yet, it became clear that forging our path, creating a company for a new age, was the direction to follow. Reflecting on how the world has transformed since I began my journey in 1984, without a computer, fax machine, or cell phone, I’m convinced that the next five years will bring about even more dramatic changes, especially with advancements in AI. At BKREA, we’re embracing this future by integrating AI technologies to interpret data in new ways and streamline our transaction processes, all while holding onto our valued analog methods, like those cherished here in the map room. Our commitment is to our clients, ensuring the best outcomes through innovation and efficiency. I’m also excited to introduce Seth Samowitz as our new COO. With a tremendous background in AI, Seth is the perfect leader to guide us in this new direction. Our goal is to assemble the best team for our clients, selecting top talent across the entire market to ensure the finest execution in every venture. Moreover, we’re launching the Knakal Affiliate Program (KAP) to strengthen connections and collaboration within our industry. I encourage you to visit our website, to learn more about how we’re transforming real estate advisory services. My cell number remains the same, and I’m eager to connect or reconnect with many of you. Feel free to reach out at BK@BKREA.com. A huge thank you again for your support and kindness. I look forward to connecting with you online or at events. Here’s to a future where tradition and technology converge to create value for our clients!
4:57

Exciting News from the Knakal Map Room Bob Knakal here, and I’m thrilled to share some exciting news about my latest venture. After a period of reflection and immense support from this incredible community, I’m proud to announce the launch of BKREA (BKREA). BKREA is born from an entrepreneurial spirit reawakened, with a vision to redefine the capital markets landscape through investment sales, debt, and equity. This wouldn't have been possible without the heartwarming texts, DMs, and voice messages from many of you. Your support during the challenging times was incredible, and I am deeply grateful for every word of encouragement. In the past few weeks, I've met with numerous individuals and companies, each with fantastic platforms and innovative ideas. Yet, it became clear that forging our path, creating a company for a new age, was the direction to follow. Reflecting on how the world has transformed since I began my journey in 1984, without a computer, fax machine, or cell phone, I’m convinced that the next five years will bring about even more dramatic changes, especially with advancements in AI. At BKREA, we’re embracing this future by integrating AI technologies to interpret data in new ways and streamline our transaction processes, all while holding onto our valued analog methods, like those cherished here in the map room. Our commitment is to our clients, ensuring the best outcomes through innovation and efficiency. I’m also excited to introduce Seth Samowitz as our new COO. With a tremendous background in AI, Seth is the perfect leader to guide us in this new direction. Our goal is to assemble the best team for our clients, selecting top talent across the entire market to ensure the finest execution in every venture. Moreover, we’re launching the Knakal Affiliate Program (KAP) to strengthen connections and collaboration within our industry. I encourage you to visit our website, to learn more about how we’re transforming real estate advisory services. My cell number remains the same, and I’m eager to connect or reconnect with many of you. Feel free to reach out at [email protected]. A huge thank you again for your support and kindness. I look forward to connecting with you online or at events. Here’s to a future where tradition and technology converge to create value for our clients!

Bob Knakal | NYC Investment Sales

181,963 görüntüleme • 2 yıl önce

🌏 Speech for Thailand GCV Industry Alliance Conference September 14, 2025 – Vayupak Grand Ballroom, Bangkok Dear Thailand GCV Ambassadors, merchants and pioneers, Dear Ms. Rung Arun Maneecho– Head GCV Ambassador, Mr. Phupanthong Chishomphu – Ecosystem Development Ambassador Mr. Banjong Wongnarong– GCV Education Ambassador Thank you very much for your kind invitation to this Thailand GCV Industry Alliance Conference. Although I cannot be present in person or live video because it is a.m. in my time zone, I feel your passion, persistence, and commitment deeply. Thailand is an amazing community. A few months ago, I saw pioneers here actively supporting GCV. I thought: How wonderful if Thailand could join the GCV Ambassador family. Soon after, Ms. Lumari told me she received Thailand’s Ambassador application. That moment felt like destiny. Since then, I have watched your leadership grow: Ms. Rung Arun Maneechot – guiding with wisdom and vision. Mr. Phupanthong Chishomphu – developing the ecosystem with energy. Mr. Banjong Wongnarong educating pioneers with dedication. Your GCV Ambassadors structure in 74 provinces and countless barter activities are shared across our network. Each time, Ms. Lumari proudly says: “Thailand community is amazing because they have amazing community leaders!” Even though you joined not long ago, your community is united, efficient, and inspiring. That is why I rarely have speech in your group—you already have great leaders who carry our strategy forward under Ms.Lumar’s guidance. Today marks another milestone: the beginning of Thailand’s GCV Industry Alliance. With many merchants joining this event, I want to explain why this alliance is so important. And big congratulation on Thailand GCV Industry Alliance has been successfully established. Why Do We Need the GCV Industry Alliance? Because this is how we bring Pi’s vision into real practice—how Pi payments benefit ordinary people and real businesses. We have done many barter activities and created total around 20 million GCV data and Thailand has created above 1 million, but their influence is limited. Merchants care about deeper questions: Can I use Pi to pay my suppliers? Will cooperation bring me more sales? Can it reduce my advertising costs and increase competitiveness? As a CPA and Tax Professional, I know why businesses fail. The main issue is not only lack of customers, but also lack of cash flow. Many companies wait months to receive payments. Some customers are never paid. This is why accountants record bad debts—a painful reality for businesses everywhere. The GCV Industry Alliance offers a solution. With partial Pi and partial fiat, merchants can cover costs in fiat immediately, while using Pi for profit. Each merchant can set their own Pi partial rate, as long as products remain competitive. When the business customer cannot pay off all the Account Receivable, but if they only need to pay partial of FIAT, they can have the ability to pay more back. And partial Pi payment can attract more business customers. After OM, all merchants will use fixed Pi to FIAT rate for full Pi payment and the merchants have the customers already. So, all merchants please make sure your FIAT can cover all cost and some partial of profit. Only partial of profit in Pi payment. This will make sure all merchants won’t incur loss. Think about it: pioneers buy groceries and essentials every day. Why not channel that demand into a system where everyone benefits? Retail chains will gain steady customers. By linking retailers to wholesalers and to suppliers, and suppliers to factories, every level of the chain can profit. That means merchants who join GCV Industry Alliance can have competitive advantages because they can reach millions of pioneers. Beyond national trade, Industry Alliances also prepare us for international trade. Each country has unique strengths. By connecting across borders, we share competitive advantages and build a stronger global economy under GCV. And most importantly, this positions us for the next stage. A strong Industry Alliance will speed our journey to Open Mainnet at GCV. After OM, it can reduce selling pressure on Pi and make conversion to fiat possible and sustainable. Closing This is why I strongly encourage merchants to join the GCV Industry Alliance. The benefits are clear: higher sales, stronger competitiveness, healthier cash flow, and long-term stability. I believe Thailand will have a very successful Industry Alliance. Finally, I wish today’s conference great success. Thank you to the Thailand GCV community for your amazing contributions. Together, we are building the foundation for Pi’s global future. Thank you. Doris Yin Founder, Global GCV Movement 9/14/2025
8:44

🌏 Speech for Thailand GCV Industry Alliance Conference September 14, 2025 – Vayupak Grand Ballroom, Bangkok Dear Thailand GCV Ambassadors, merchants and pioneers, Dear Ms. Rung Arun Maneecho– Head GCV Ambassador, Mr. Phupanthong Chishomphu – Ecosystem Development Ambassador Mr. Banjong Wongnarong– GCV Education Ambassador Thank you very much for your kind invitation to this Thailand GCV Industry Alliance Conference. Although I cannot be present in person or live video because it is a.m. in my time zone, I feel your passion, persistence, and commitment deeply. Thailand is an amazing community. A few months ago, I saw pioneers here actively supporting GCV. I thought: How wonderful if Thailand could join the GCV Ambassador family. Soon after, Ms. Lumari told me she received Thailand’s Ambassador application. That moment felt like destiny. Since then, I have watched your leadership grow: Ms. Rung Arun Maneechot – guiding with wisdom and vision. Mr. Phupanthong Chishomphu – developing the ecosystem with energy. Mr. Banjong Wongnarong educating pioneers with dedication. Your GCV Ambassadors structure in 74 provinces and countless barter activities are shared across our network. Each time, Ms. Lumari proudly says: “Thailand community is amazing because they have amazing community leaders!” Even though you joined not long ago, your community is united, efficient, and inspiring. That is why I rarely have speech in your group—you already have great leaders who carry our strategy forward under Ms.Lumar’s guidance. Today marks another milestone: the beginning of Thailand’s GCV Industry Alliance. With many merchants joining this event, I want to explain why this alliance is so important. And big congratulation on Thailand GCV Industry Alliance has been successfully established. Why Do We Need the GCV Industry Alliance? Because this is how we bring Pi’s vision into real practice—how Pi payments benefit ordinary people and real businesses. We have done many barter activities and created total around 20 million GCV data and Thailand has created above 1 million, but their influence is limited. Merchants care about deeper questions: Can I use Pi to pay my suppliers? Will cooperation bring me more sales? Can it reduce my advertising costs and increase competitiveness? As a CPA and Tax Professional, I know why businesses fail. The main issue is not only lack of customers, but also lack of cash flow. Many companies wait months to receive payments. Some customers are never paid. This is why accountants record bad debts—a painful reality for businesses everywhere. The GCV Industry Alliance offers a solution. With partial Pi and partial fiat, merchants can cover costs in fiat immediately, while using Pi for profit. Each merchant can set their own Pi partial rate, as long as products remain competitive. When the business customer cannot pay off all the Account Receivable, but if they only need to pay partial of FIAT, they can have the ability to pay more back. And partial Pi payment can attract more business customers. After OM, all merchants will use fixed Pi to FIAT rate for full Pi payment and the merchants have the customers already. So, all merchants please make sure your FIAT can cover all cost and some partial of profit. Only partial of profit in Pi payment. This will make sure all merchants won’t incur loss. Think about it: pioneers buy groceries and essentials every day. Why not channel that demand into a system where everyone benefits? Retail chains will gain steady customers. By linking retailers to wholesalers and to suppliers, and suppliers to factories, every level of the chain can profit. That means merchants who join GCV Industry Alliance can have competitive advantages because they can reach millions of pioneers. Beyond national trade, Industry Alliances also prepare us for international trade. Each country has unique strengths. By connecting across borders, we share competitive advantages and build a stronger global economy under GCV. And most importantly, this positions us for the next stage. A strong Industry Alliance will speed our journey to Open Mainnet at GCV. After OM, it can reduce selling pressure on Pi and make conversion to fiat possible and sustainable. Closing This is why I strongly encourage merchants to join the GCV Industry Alliance. The benefits are clear: higher sales, stronger competitiveness, healthier cash flow, and long-term stability. I believe Thailand will have a very successful Industry Alliance. Finally, I wish today’s conference great success. Thank you to the Thailand GCV community for your amazing contributions. Together, we are building the foundation for Pi’s global future. Thank you. Doris Yin Founder, Global GCV Movement 9/14/2025

Doris Yin 东方紫莲🪷

18,244 görüntüleme • 11 ay önce

How to Improve Your Sex Life & Strengthen Your Brain 💪🧠 We're diving deep into the secrets of sexual wellness with board-certified urologist Dr. Rena Malik (Rena Malik, MD | Urologist)! From penis size myths to the mysteries of pelvic floor health, we're covering everything you've ever wanted to know about intimacy and sexual wellness. Dr. Malik explains why some of us aren’t having the orgasms we crave & how to change that, how sex can boost your brain & overall health, who is most likely to have the healthiest sex life, practical tips to boost libido, and the challenges of navigating intimacy in today’s hookup culture. She also breaks down 🌟: - Pain During Sex?: Discover the causes and solutions that could change your life! - Intimacy and Orgasm: Why emotional connection is the secret sauce for pleasure - Porn's Impact on Real Life: Dangers of copying what you see - Perimenopause and Menopause: Symptoms, treatments, and what to expect - Top Myths About Female Orgasms - Scientific Proof for Why Size Doesn't Matter 🍆 - What NOT to Say When He Can't Get It Up: Essential advice for sensitive conversations around ED - Why more women are exploring same-sex relationships 🌈 - How often should couples really be getting it on? - Health Benefits & Risks of Semen Retention - Proper way to do Kegel Exercises for maximum benefits - Why lack of orgasms in sex is NOT a personal failure - Benefits & Downsides of Sex Toys in the Bedroom - Is the G-Spot Real? - How our sex lives change as we age Find out how breaking the stigma around sexual discussions can change your life - TUNE IN to MBB to transform your understanding of sex, intimacy, and health today! 🔥✨ Want to see the complete episode? Click the ▶️ 🧠💥 #BialikBreakdown #mayimbialik #mentalhealthpodcast #podcast
0:56

How to Improve Your Sex Life & Strengthen Your Brain 💪🧠 We're diving deep into the secrets of sexual wellness with board-certified urologist Dr. Rena Malik (Rena Malik, MD | Urologist)! From penis size myths to the mysteries of pelvic floor health, we're covering everything you've ever wanted to know about intimacy and sexual wellness. Dr. Malik explains why some of us aren’t having the orgasms we crave & how to change that, how sex can boost your brain & overall health, who is most likely to have the healthiest sex life, practical tips to boost libido, and the challenges of navigating intimacy in today’s hookup culture. She also breaks down 🌟: - Pain During Sex?: Discover the causes and solutions that could change your life! - Intimacy and Orgasm: Why emotional connection is the secret sauce for pleasure - Porn's Impact on Real Life: Dangers of copying what you see - Perimenopause and Menopause: Symptoms, treatments, and what to expect - Top Myths About Female Orgasms - Scientific Proof for Why Size Doesn't Matter 🍆 - What NOT to Say When He Can't Get It Up: Essential advice for sensitive conversations around ED - Why more women are exploring same-sex relationships 🌈 - How often should couples really be getting it on? - Health Benefits & Risks of Semen Retention - Proper way to do Kegel Exercises for maximum benefits - Why lack of orgasms in sex is NOT a personal failure - Benefits & Downsides of Sex Toys in the Bedroom - Is the G-Spot Real? - How our sex lives change as we age Find out how breaking the stigma around sexual discussions can change your life - TUNE IN to MBB to transform your understanding of sex, intimacy, and health today! 🔥✨ Want to see the complete episode? Click the ▶️ 🧠💥 #BialikBreakdown #mayimbialik #mentalhealthpodcast #podcast

Mayim Bialik

18,175 görüntüleme • 1 yıl önce

⚡️We are excited to announce that our new no-code Enterprise Platform is NOW available in private beta! As RAG apps advance from prototype to production we’ve been overwhelmed by requests for an enterprise grade solution to provide these applications with the data they need. Designed to make it easy to get your data #RAGready, our Platform can preprocess more than 25 file types and soon will be fully #multimodal, also able to ingest audio, video and image files. We ship with a baseline suite of source connectors, including Amazon Web Services S3, Microsoft Azure Blob Storage, OneDrive, SFTP, Databricks Delta Table, Google Drive, Salesforce, Elastic, OpenSearch, and Google Cloud storage with many more fast following. Platform transforms your documents into a standardized JSON schema, broken down into semantically coherent elements allowing you to reconstruct your document in the manner most useful to you. Want only the narrative text but not the headers and footers? This is entirely configurable through the UI. Additionally, we generate more than 30 types of metadata for each element to make it easy to curate the data being written downstream and to support metadata filtering during retrieval. Smart chunking and the ability to choose from a range of embedding models are in from launch, delivering a turnkey solution for chunk and embedding experimentation. As for destination connectors, we've got that covered too, with Amazon Web Services S3, Pinecone, Chroma , Weaviate AI Database, Google Cloud storage, MongoDB, Microsoft Azure cognitive search, PostgreSQL, Elastic, OpenSearch, and Databricks Delta Table. And of course, all of this can be scheduled to keep your data continuously hydrated. The private-beta is live today! Sign-up to get access and come build the future of LLM data foundations with us: 🚀 #ETLforLLMs #AI #DataPreprocessing #DataScience #DataTransformation #LLMs #ETL #ML #PreppingData #MachineLearning #RAG #Engineer #Unstructured #Unstructuredio #RetrievalAugmentedGeneration #multimodal #AIJobs
1:00

⚡️We are excited to announce that our new no-code Enterprise Platform is NOW available in private beta! As RAG apps advance from prototype to production we’ve been overwhelmed by requests for an enterprise grade solution to provide these applications with the data they need. Designed to make it easy to get your data #RAGready, our Platform can preprocess more than 25 file types and soon will be fully #multimodal, also able to ingest audio, video and image files. We ship with a baseline suite of source connectors, including Amazon Web Services S3, Microsoft Azure Blob Storage, OneDrive, SFTP, Databricks Delta Table, Google Drive, Salesforce, Elastic, OpenSearch, and Google Cloud storage with many more fast following. Platform transforms your documents into a standardized JSON schema, broken down into semantically coherent elements allowing you to reconstruct your document in the manner most useful to you. Want only the narrative text but not the headers and footers? This is entirely configurable through the UI. Additionally, we generate more than 30 types of metadata for each element to make it easy to curate the data being written downstream and to support metadata filtering during retrieval. Smart chunking and the ability to choose from a range of embedding models are in from launch, delivering a turnkey solution for chunk and embedding experimentation. As for destination connectors, we've got that covered too, with Amazon Web Services S3, Pinecone, Chroma , Weaviate AI Database, Google Cloud storage, MongoDB, Microsoft Azure cognitive search, PostgreSQL, Elastic, OpenSearch, and Databricks Delta Table. And of course, all of this can be scheduled to keep your data continuously hydrated. The private-beta is live today! Sign-up to get access and come build the future of LLM data foundations with us: 🚀 #ETLforLLMs #AI #DataPreprocessing #DataScience #DataTransformation #LLMs #ETL #ML #PreppingData #MachineLearning #RAG #Engineer #Unstructured #Unstructuredio #RetrievalAugmentedGeneration #multimodal #AIJobs

Unstructured

21,874 görüntüleme • 2 yıl önce

As I promised yesterday, I'll briefly explain LoRA training and share a workflow I made so you can do it quickly. First, let me answer a very common question: 'Why train LoRAs when we have such advanced models?' Even though we have incredibly advanced models now (like NBP), we still can't always get them to do specific things we want. Simplest example: the spritesheet LoRA I made the other day. I generated 1000 images with Nano Banana and only 100 were what I wanted. The LoRA I trained using those 100 images gives me nearly 100% consistent results. Second point is cost and speed. With LoRA, we can cut costs by 4-5x. And while doing that, we're generating 4-5x faster. How many images do you need for a good LoRA? This depends on your LoRA's complexity. For example, when I training the spritesheet LoRA, even though I used 100 images, I didn't include buildings in the training data, so this LoRA doesn't work for buildings. So think about your LoRA's use cases and add examples for as many use cases as possible to improve quality. What are paired images and how to train LoRAs for image-editing? When training LoRAs for image editing on fal, we call each edit example paired images - one with _start suffix, one with _end suffix. For example, if you're training a background remove LoRA, the unedited original photo will be your '_start' image. The image with background removed will be the '_end' image. Simply put: images we want to edit or use as reference get _start, target images we want to achieve get '_end'. Important: save both images with the same name. Like image332_start.jpg and image332_end.jpg. This way the system knows which images pair together. What about training LoRAs for models with multiple image inputs? Same logic. We still use _start and _end suffixes, but with one difference. Since there are multiple input images, we can number them: _start, _start1, _start2. Example: start images, 1st image = Woman portrait (image35_start.jpg) 2nd image = Glasses photo (image35_start1.jpg) 3rd image = Hat photo (image35_start2.jpg) Output image = portrait of woman wearing glasses and hat (image35_end.jpg) Can we do more detailed captioning? Yes. Similarly, you can improve training quality by creating a txt file for each set with the caption inside. Example: create image35.txt and write: 'Recreate the image by putting the glasses from the second image and the hat from the third image on the woman in the first image.' What are Steps? How many should I use? What's Learning Rate? Steps determines how many times the model sees and processes your training data (your images). Each step, the model learns a bit more. But as steps increase, so does the risk of overfitting. So there's no real default. But for a simpler LoRA with 20 paired images, 1000 steps is ideal. Here's a metaphor for the Steps and Learning Rate relationship: Imagine you have a balloon. Our goal is to inflate it to the optimal size. Steps = How many times we blow into the balloon Learning rate = How hard we blow each time If we blow too softly, we need to blow many more times. If we blow too hard, we risk popping it quickly and can't reach optimal size. Of course training won't explode, but it won't work as intended because it wasn't trained optimally. Training's done, now what? Once training's complete, you'll have a safetensors file. Every model you train on fal has a LoRA inference endpoint. In that inference, add your safetensors file link to the LoRA url input, and you can use your LoRA. Thanks for the read! The workflow in the video: If I forgot anything, let me know in the replies.
1:06

As I promised yesterday, I'll briefly explain LoRA training and share a workflow I made so you can do it quickly. First, let me answer a very common question: 'Why train LoRAs when we have such advanced models?' Even though we have incredibly advanced models now (like NBP), we still can't always get them to do specific things we want. Simplest example: the spritesheet LoRA I made the other day. I generated 1000 images with Nano Banana and only 100 were what I wanted. The LoRA I trained using those 100 images gives me nearly 100% consistent results. Second point is cost and speed. With LoRA, we can cut costs by 4-5x. And while doing that, we're generating 4-5x faster. How many images do you need for a good LoRA? This depends on your LoRA's complexity. For example, when I training the spritesheet LoRA, even though I used 100 images, I didn't include buildings in the training data, so this LoRA doesn't work for buildings. So think about your LoRA's use cases and add examples for as many use cases as possible to improve quality. What are paired images and how to train LoRAs for image-editing? When training LoRAs for image editing on fal, we call each edit example paired images - one with _start suffix, one with _end suffix. For example, if you're training a background remove LoRA, the unedited original photo will be your '_start' image. The image with background removed will be the '_end' image. Simply put: images we want to edit or use as reference get _start, target images we want to achieve get '_end'. Important: save both images with the same name. Like image332_start.jpg and image332_end.jpg. This way the system knows which images pair together. What about training LoRAs for models with multiple image inputs? Same logic. We still use _start and _end suffixes, but with one difference. Since there are multiple input images, we can number them: _start, _start1, _start2. Example: start images, 1st image = Woman portrait (image35_start.jpg) 2nd image = Glasses photo (image35_start1.jpg) 3rd image = Hat photo (image35_start2.jpg) Output image = portrait of woman wearing glasses and hat (image35_end.jpg) Can we do more detailed captioning? Yes. Similarly, you can improve training quality by creating a txt file for each set with the caption inside. Example: create image35.txt and write: 'Recreate the image by putting the glasses from the second image and the hat from the third image on the woman in the first image.' What are Steps? How many should I use? What's Learning Rate? Steps determines how many times the model sees and processes your training data (your images). Each step, the model learns a bit more. But as steps increase, so does the risk of overfitting. So there's no real default. But for a simpler LoRA with 20 paired images, 1000 steps is ideal. Here's a metaphor for the Steps and Learning Rate relationship: Imagine you have a balloon. Our goal is to inflate it to the optimal size. Steps = How many times we blow into the balloon Learning rate = How hard we blow each time If we blow too softly, we need to blow many more times. If we blow too hard, we risk popping it quickly and can't reach optimal size. Of course training won't explode, but it won't work as intended because it wasn't trained optimally. Training's done, now what? Once training's complete, you'll have a safetensors file. Every model you train on fal has a LoRA inference endpoint. In that inference, add your safetensors file link to the LoRA url input, and you can use your LoRA. Thanks for the read! The workflow in the video: If I forgot anything, let me know in the replies.

ilker

15,184 görüntüleme • 7 ay önce