☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 days agoA good primer on what to expect running local LLMsplus-squarenullprogram.comexternal-linkmessage-square4fedilinkarrow-up13arrow-down10
arrow-up13arrow-down1external-linkA good primer on what to expect running local LLMsplus-squarenullprogram.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 days agomessage-square4fedilink
Shamar@feddit.itEnglish · edit-224 days agoA community statement supporting the Open Source Definition (OSD)plus-squareosd.fyiexternal-linkmessage-square1fedilinkarrow-up111arrow-down11
arrow-up110arrow-down1external-linkA community statement supporting the Open Source Definition (OSD)plus-squareosd.fyiShamar@feddit.itEnglish · edit-224 days agomessage-square1fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 2 months agoHow ‘Embeddings’ Encode What Words Meanplus-squarewww.quantamagazine.orgexternal-linkmessage-square0fedilinkarrow-up19arrow-down10
arrow-up19arrow-down1external-linkHow ‘Embeddings’ Encode What Words Meanplus-squarewww.quantamagazine.org☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 2 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoNew AI model “learns” how to simulate Super Mario Bros. from video footageplus-squarearstechnica.comexternal-linkmessage-square0fedilinkarrow-up14arrow-down10
arrow-up14arrow-down1external-linkNew AI model “learns” how to simulate Super Mario Bros. from video footageplus-squarearstechnica.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoReflection 70B holds its own against even the top closed-source models (Claude 3.5 Sonnet, GPT-4o)plus-squarehuggingface.coexternal-linkmessage-square0fedilinkarrow-up112arrow-down10
arrow-up112arrow-down1external-linkReflection 70B holds its own against even the top closed-source models (Claude 3.5 Sonnet, GPT-4o)plus-squarehuggingface.co☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoIt’s Not Intelligent If It Always Halts: A Critical Perspective on Current Approaches to AGIplus-squarewww.lifeiscomputation.comexternal-linkmessage-square6fedilinkarrow-up19arrow-down10
arrow-up19arrow-down1external-linkIt’s Not Intelligent If It Always Halts: A Critical Perspective on Current Approaches to AGIplus-squarewww.lifeiscomputation.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square6fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoThe Difference Between Speaking and Thinkingplus-squarewww.theatlantic.comexternal-linkmessage-square0fedilinkarrow-up16arrow-down10
arrow-up16arrow-down1external-linkThe Difference Between Speaking and Thinkingplus-squarewww.theatlantic.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoDiffusion Models Are Real-Time Game Enginesplus-squaregamengen.github.ioexternal-linkmessage-square0fedilinkarrow-up16arrow-down10
arrow-up16arrow-down1external-linkDiffusion Models Are Real-Time Game Enginesplus-squaregamengen.github.io☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoLiger Kernel is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%.plus-squaregithub.comexternal-linkmessage-square0fedilinkarrow-up14arrow-down10
arrow-up14arrow-down1external-linkLiger Kernel is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%.plus-squaregithub.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoTransformer Explainerplus-squarepoloclub.github.ioexternal-linkmessage-square0fedilinkarrow-up12arrow-down10
arrow-up12arrow-down1external-linkTransformer Explainerplus-squarepoloclub.github.io☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agoAlibaba claims no. 1 spot in AI math models with Qwen2-Mathventurebeat.comexternal-linkmessage-square0fedilinkarrow-up19arrow-down11
arrow-up18arrow-down1external-linkAlibaba claims no. 1 spot in AI math models with Qwen2-Mathventurebeat.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 3 months agomessage-square0fedilink
yboutros@infosec.pubEnglish · 3 months agoHow to convert a positionally encoded predicted embedding from a decoder to its matching token?plus-squaremessage-squaremessage-square2fedilinkarrow-up14arrow-down10
arrow-up14arrow-down1message-squareHow to convert a positionally encoded predicted embedding from a decoder to its matching token?plus-squareyboutros@infosec.pubEnglish · 3 months agomessage-square2fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 4 months agoNew Open-Source AI Image Generator Beats Midjourney, SD3 and Auraflowplus-squaredecrypt.coexternal-linkmessage-square1fedilinkarrow-up111arrow-down10
arrow-up111arrow-down1external-linkNew Open-Source AI Image Generator Beats Midjourney, SD3 and Auraflowplus-squaredecrypt.co☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 4 months agomessage-square1fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 4 months agoAI models collapse when trained on recursively generated dataplus-squarewww.nature.comexternal-linkmessage-square3fedilinkarrow-up128arrow-down11
arrow-up127arrow-down1external-linkAI models collapse when trained on recursively generated dataplus-squarewww.nature.com☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 4 months agomessage-square3fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agoRouteLLM: An Open-Source Framework for Cost-Effective LLM Routingplus-squarelmsys.orgexternal-linkmessage-square0fedilinkarrow-up14arrow-down13
arrow-up11arrow-down1external-linkRouteLLM: An Open-Source Framework for Cost-Effective LLM Routingplus-squarelmsys.org☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agoAlibaba's Qwen LLM model leading open source rankingsplus-squarehuggingface.coexternal-linkmessage-square0fedilinkarrow-up16arrow-down10
arrow-up16arrow-down1external-linkAlibaba's Qwen LLM model leading open source rankingsplus-squarehuggingface.co☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agomessage-square0fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · edit-25 months agoBy using the same techniques Google used to solve Go (MTCS and backprop), Llama8B gets 96.7% on math benchmark GSM8K. That’s better than GPT-4, Claude and Gemini, with 200x fewer parameters!plus-squarearxiv.orgexternal-linkmessage-square2fedilinkarrow-up111arrow-down13
arrow-up18arrow-down1external-linkBy using the same techniques Google used to solve Go (MTCS and backprop), Llama8B gets 96.7% on math benchmark GSM8K. That’s better than GPT-4, Claude and Gemini, with 200x fewer parameters!plus-squarearxiv.org☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · edit-25 months agomessage-square2fedilink
☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agoMixture of Agents (MoA) leverages several open-source LLM agents to achieve a score of 65.1% on AlpacaEval 2.0plus-squarewww.together.aiexternal-linkmessage-square0fedilinkarrow-up17arrow-down11
arrow-up16arrow-down1external-linkMixture of Agents (MoA) leverages several open-source LLM agents to achieve a score of 65.1% on AlpacaEval 2.0plus-squarewww.together.ai☆ Yσɠƚԋσʂ ☆@lemmy.mlEnglish · 5 months agomessage-square0fedilink
ylai@lemmy.mlEnglish · 5 months agoFrom DeepSpeed to FSDP and Back Again with Hugging Face Accelerateplus-squarehuggingface.coexternal-linkmessage-square0fedilinkarrow-up14arrow-down10
arrow-up14arrow-down1external-linkFrom DeepSpeed to FSDP and Back Again with Hugging Face Accelerateplus-squarehuggingface.coylai@lemmy.mlEnglish · 5 months agomessage-square0fedilink
keepthepace@slrpnk.net · 5 months agoTorrent tracker for open modelsplus-squareaitracker.artexternal-linkmessage-square0fedilinkarrow-up19arrow-down11
arrow-up18arrow-down1external-linkTorrent tracker for open modelsplus-squareaitracker.artkeepthepace@slrpnk.net · 5 months agomessage-square0fedilink
