#nlp — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #nlp, aggregated by home.social.
-
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
Five frontier LLMs disagree on 67% of 1k real-world fact-check claims
https://lenz.io/research/llm-disagreement
#HackerNews #LLMs #AIresearch #factchecking #disagreements #NLP
-
#DuckDuckGo saw a 30% increase in U.S. app installs and a 22.7% increase in visits to its AI-free search page following #Google’s announcement of its #AI driven search overhaul. Users are concerned about Google’s #AIintegration, citing issues with accuracy and control. DuckDuckGo emphasises user choice and #privacy, offering both AI and non-AI search options. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#DuckDuckGo saw a 30% increase in U.S. app installs and a 22.7% increase in visits to its AI-free search page following #Google’s announcement of its #AI driven search overhaul. Users are concerned about Google’s #AIintegration, citing issues with accuracy and control. DuckDuckGo emphasises user choice and #privacy, offering both AI and non-AI search options. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#DuckDuckGo saw a 30% increase in U.S. app installs and a 22.7% increase in visits to its AI-free search page following #Google’s announcement of its #AI driven search overhaul. Users are concerned about Google’s #AIintegration, citing issues with accuracy and control. DuckDuckGo emphasises user choice and #privacy, offering both AI and non-AI search options. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#DuckDuckGo saw a 30% increase in U.S. app installs and a 22.7% increase in visits to its AI-free search page following #Google’s announcement of its #AI driven search overhaul. Users are concerned about Google’s #AIintegration, citing issues with accuracy and control. DuckDuckGo emphasises user choice and #privacy, offering both AI and non-AI search options. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#DuckDuckGo saw a 30% increase in U.S. app installs and a 22.7% increase in visits to its AI-free search page following #Google’s announcement of its #AI driven search overhaul. Users are concerned about Google’s #AIintegration, citing issues with accuracy and control. DuckDuckGo emphasises user choice and #privacy, offering both AI and non-AI search options. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
أهلاً معجبيّ التقنية والخصوصية!
المقال يوضح كيف أن نماذج BERT وGPT تُحدث ثورة في معالجة اللغة الطبيعية 🌐- BERT: فهم ثنائي الاتجاه = تمكين الترجمة الدقيقة وتحليل المشاعر مع الحفاظ على الخصوصية
- GPT: توليد نص طبيعي ومتعدد الاستخدامات = إنشاء محتوى موثوق وتطابق السياقتأثيرها: أبحاث أكاديمية ثرية، أدوات ثورية، شفافية أكبر في تعاملات الذكاء الاصطناعي.
-
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
#GoogleCloud COO #FrancisdeSouza emphasises the need for a #platformapproach to #AIsecurity, integrating #security from the start and addressing the threat of #shadowAI. He highlights the need for consistent security across clouds and the emergence of AI-native defences. https://techcrunch.com/2026/05/24/everyone-is-navigating-ai-security-in-real-time-even-google/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#GoogleCloud COO #FrancisdeSouza emphasises the need for a #platformapproach to #AIsecurity, integrating #security from the start and addressing the threat of #shadowAI. He highlights the need for consistent security across clouds and the emergence of AI-native defences. https://techcrunch.com/2026/05/24/everyone-is-navigating-ai-security-in-real-time-even-google/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#GoogleCloud COO #FrancisdeSouza emphasises the need for a #platformapproach to #AIsecurity, integrating #security from the start and addressing the threat of #shadowAI. He highlights the need for consistent security across clouds and the emergence of AI-native defences. https://techcrunch.com/2026/05/24/everyone-is-navigating-ai-security-in-real-time-even-google/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#GoogleCloud COO #FrancisdeSouza emphasises the need for a #platformapproach to #AIsecurity, integrating #security from the start and addressing the threat of #shadowAI. He highlights the need for consistent security across clouds and the emergence of AI-native defences. https://techcrunch.com/2026/05/24/everyone-is-navigating-ai-security-in-real-time-even-google/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#GoogleCloud COO #FrancisdeSouza emphasises the need for a #platformapproach to #AIsecurity, integrating #security from the start and addressing the threat of #shadowAI. He highlights the need for consistent security across clouds and the emergence of AI-native defences. https://techcrunch.com/2026/05/24/everyone-is-navigating-ai-security-in-real-time-even-google/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - AI Can Mention the Oppressed and Still Strip Them of Agency
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://app.hackernoon.com/mobile/6a0b1298e3fb338333ad92d1
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
How Can We Prevent AI Models From Cannibalizing Themselves When Human-Generated Data Runs Out?
Getty Images While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from. Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse."......Continue reading... By: Roland Moore-Colyer Source: Live Science . Critics: A backdoor in a […] -
How Can We Prevent AI Models From Cannibalizing Themselves When Human-Generated Data Runs Out?
Getty Images While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from. Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse."......Continue reading... By: Roland Moore-Colyer Source: Live Science . Critics: A backdoor in a […] -
How Can We Prevent AI Models From Cannibalizing Themselves When Human-Generated Data Runs Out?
Getty Images While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from. Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse."......Continue reading... By: Roland Moore-Colyer Source: Live Science . Critics: A backdoor in a […] -
How Can We Prevent AI Models From Cannibalizing Themselves When Human-Generated Data Runs Out?
Getty Images While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from. Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse."......Continue reading... By: Roland Moore-Colyer Source: Live Science . Critics: A backdoor in a […] -
How Can We Prevent AI Models From Cannibalizing Themselves When Human-Generated Data Runs Out?
Getty Images While the evolution of artificial intelligence (AI) systems has shown no sign of slowing, there's a growing concern that large language models (LLMs) will soon run out of human-made data to ingest and learn from. Once this happens, scientists say, AI models will increasingly rely on synthetic AI-made information, which will lead to an effect called "model collapse."......Continue reading... By: Roland Moore-Colyer Source: Live Science . Critics: A backdoor in a […] -
🚨 New Article - Suffering Without Perpetrators: The Humanitarian Passive in AI-Generated Conflict Discourse
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6753123
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - Suffering Without Perpetrators: The Humanitarian Passive in AI-Generated Conflict Discourse
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6753123
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - Suffering Without Perpetrators: The Humanitarian Passive in AI-Generated Conflict Discourse
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6753123
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - Suffering Without Perpetrators: The Humanitarian Passive in AI-Generated Conflict Discourse
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6753123
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
🚨 New Article - Suffering Without Perpetrators: The Humanitarian Passive in AI-Generated Conflict Discourse
Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency.
🔗https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6753123
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg
#healthcare #ArtificialIntelligence #NLP #aifutures #lawstodon
#tech #agustinvstartari #linguistics #ai #LRM -
[Перевод] Нехватка CUDA-памяти при обучении с GRPO: как перестать гадать и начать считать
Ошибка CUDA out of memory при обучении LLM обычно превращается в бесконечный цикл случайных правок: уменьшили batch size, урезали sequence length, снизили LoRA rank — и всё равно где-то снова падает. Особенно весело становится в RL-сценариях с GRPO, vLLM и генерацией нескольких ответов на один промпт. Поговорим о том, как перестать гадать и начать считать потребление GPU-памяти: от чтения самого текста ошибки до оценки вклада vLLM, активаций и параметров обучения. С формулами, реальными конфигами и объяснением, какие настройки действительно дают эффект, а какие только создают иллюзию оптимизации. Оптимизировать LLM
https://habr.com/ru/companies/otus/articles/1037332/
#NLP #LLM #GRPO #обучение_с_подкреплением #CUDA_out_of_memory #vLLM #оптимизация_GPU_памяти #дообучение_моделей #LoRA #PyTorch
-
[Перевод] Нехватка CUDA-памяти при обучении с GRPO: как перестать гадать и начать считать
Ошибка CUDA out of memory при обучении LLM обычно превращается в бесконечный цикл случайных правок: уменьшили batch size, урезали sequence length, снизили LoRA rank — и всё равно где-то снова падает. Особенно весело становится в RL-сценариях с GRPO, vLLM и генерацией нескольких ответов на один промпт. Поговорим о том, как перестать гадать и начать считать потребление GPU-памяти: от чтения самого текста ошибки до оценки вклада vLLM, активаций и параметров обучения. С формулами, реальными конфигами и объяснением, какие настройки действительно дают эффект, а какие только создают иллюзию оптимизации. Оптимизировать LLM
https://habr.com/ru/companies/otus/articles/1037332/
#NLP #LLM #GRPO #обучение_с_подкреплением #CUDA_out_of_memory #vLLM #оптимизация_GPU_памяти #дообучение_моделей #LoRA #PyTorch
-
[Перевод] Нехватка CUDA-памяти при обучении с GRPO: как перестать гадать и начать считать
Ошибка CUDA out of memory при обучении LLM обычно превращается в бесконечный цикл случайных правок: уменьшили batch size, урезали sequence length, снизили LoRA rank — и всё равно где-то снова падает. Особенно весело становится в RL-сценариях с GRPO, vLLM и генерацией нескольких ответов на один промпт. Поговорим о том, как перестать гадать и начать считать потребление GPU-памяти: от чтения самого текста ошибки до оценки вклада vLLM, активаций и параметров обучения. С формулами, реальными конфигами и объяснением, какие настройки действительно дают эффект, а какие только создают иллюзию оптимизации. Оптимизировать LLM
https://habr.com/ru/companies/otus/articles/1037332/
#NLP #LLM #GRPO #обучение_с_подкреплением #CUDA_out_of_memory #vLLM #оптимизация_GPU_памяти #дообучение_моделей #LoRA #PyTorch
-
[Перевод] Нехватка CUDA-памяти при обучении с GRPO: как перестать гадать и начать считать
Ошибка CUDA out of memory при обучении LLM обычно превращается в бесконечный цикл случайных правок: уменьшили batch size, урезали sequence length, снизили LoRA rank — и всё равно где-то снова падает. Особенно весело становится в RL-сценариях с GRPO, vLLM и генерацией нескольких ответов на один промпт. Поговорим о том, как перестать гадать и начать считать потребление GPU-памяти: от чтения самого текста ошибки до оценки вклада vLLM, активаций и параметров обучения. С формулами, реальными конфигами и объяснением, какие настройки действительно дают эффект, а какие только создают иллюзию оптимизации. Оптимизировать LLM
https://habr.com/ru/companies/otus/articles/1037332/
#NLP #LLM #GRPO #обучение_с_подкреплением #CUDA_out_of_memory #vLLM #оптимизация_GPU_памяти #дообучение_моделей #LoRA #PyTorch
-
Learn more about Phu and his work: https://phusroyal.github.io/
Welcome to the team, Phu! 👋
#UKPLab #TUDarmstadt #MBZUAI #NLP #NLProc #MechanisticInterpretability #LLMs #AIInterpretability
-
Learn more about Phu and his work: https://phusroyal.github.io/
Welcome to the team, Phu! 👋
#UKPLab #TUDarmstadt #MBZUAI #NLP #NLProc #MechanisticInterpretability #LLMs #AIInterpretability
-
Learn more about Phu and his work: https://phusroyal.github.io/
Welcome to the team, Phu! 👋
#UKPLab #TUDarmstadt #MBZUAI #NLP #NLProc #MechanisticInterpretability #LLMs #AIInterpretability
