[{"data":1,"prerenderedAt":320},["ShallowReactive",2],{"project-35344787":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":28,"sentiment":29,"stance_mix":30,"scored_comments":35,"aspects":36,"strip":41,"post_text":44,"what":45,"said":75,"maker_in_thread":187,"shape":189,"grid":193,"attempts":285,"same_week":286,"maker_shows":33,"maker_hidden":312,"reader_version":313,"card":314},"35344787","Lit-LLaMA","A fully open-source (Apache 2.0)implementation of llama","title","Show HN: A fully open-source (Apache 2.0)implementation of llama","https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flit-llama","github.com","osurits",1680024788,2023,158,52,385,3109,13,"notable","ai","llm implementation","library",[24,25,26,27],"llama","open source llm","apache 2.0","llm inference","mixed",-0.337,{"positive":31,"neutral":32,"mixed":33,"critical":34},2,8,1,9,20,{"idea":37,"execution":38,"performance":40},{"positive":33,"critical":32},{"positive":39,"critical":39},0,{"positive":33,"critical":39},[39,39,39,39,39,39,39,33,33,39,42,42,42,42,39,33,33,39,39,39,33,39,39,39,39,33,33,42,42,42,42,33,33,39,43,43,42,39,39],3,5,"We believe that AI should be fully open source and part of the collective knowledge.\n\nThe original LLaMA code is GPL licensed which means any project using it must also be released under GPL.\n\nThis \"taints\" any other code and prevents meaningful academic and commercial use.\n\nLit-LLaMA solves that for good.",{"problem":46,"audience":52,"pricing":52,"motivation":53,"time_to_build":52,"stack":57,"openness":73,"stage":74,"team_size":74},{"value":47,"source":48,"evidence":49},"Original LLaMA code is GPL licensed, which restricts meaningful academic and commercial use","maker",[50],{"id":51},"post",null,{"value":54,"source":48,"evidence":55},"Make AI fully open source and avoid GPL restrictions on academic and commercial use",[56],{"id":51},[58,66],{"value":59,"source":60,"evidence":61},"PyTorch","commenter",[62],{"id":63,"by":64,"text":65},"35347299","UncleOxidant","Yeah, this is weird, because there are plenty of open source implementations of the lama model on github - alpaca.cpp in c++ is one, there are many others in PyTorch such as the…",{"value":67,"source":60,"evidence":68},"nanoGPT",[69],{"id":70,"by":71,"text":72},"35347073","lantiga","Correct, the original is GPL 3. To produce this implementation from the LLaMA paper we started from github.com\u002Fkarpathy\u002FnanoGPT, the LLaMA architecture is really similar to GPT.…","open_source","unknown",{"praise":76,"objections":91,"requests":136,"questions":151,"alternatives":165},[77,84],{"value":78,"source":60,"evidence":79},"Apache 2.0 licensing and ability to run on 8 GB VRAM",[80],{"id":81,"by":82,"text":83},"35344874","theaniketmaurya","I am in love with this implementation considering the ability to run on 8 GB VRAM and Apache 2.0 license.",{"value":85,"source":60,"evidence":86},"Seen as a win for the AI community",[87],{"id":88,"by":89,"text":90},"35345203","yewnork","I see this as a win for the AI community. The key for LLMs is to enable people to train collaboratively and innovate more quickly in this space. Are there any examples or demos…",[92,107,122,129],{"value":93,"source":60,"evidence":94},"Underlying LLaMA model weights remain research-only, not commercially licensed",[95,99,103],{"id":96,"by":97,"text":98},"35346654","cuuupid","I think OP mischaracterized the issue with the license, its more that the weights don’t fall under the same scope. They’re research use only, no commercial use allowed.",{"id":100,"by":101,"text":102},"35346455","nynx","There are already a million ways to run LLaMA. This doesn't change the issue at all, which is that the weights aren't commercially licensed.",{"id":104,"by":105,"text":106},"35346132","barefeg","But aren’t the weights still not for commercial use?",{"value":108,"source":60,"evidence":109},"Post mischaracterizes GPL as preventing open source or academic use",[110,114,118],{"id":111,"by":112,"text":113},"35346609","adeon","I think implying that GPL is not \"fully open source\" is a hot take. It's specifically designed to ensure you and anyone you distribute your code gets the same freedoms. Maybe you…",{"id":115,"by":116,"text":117},"35345635","2Gkashmiri","Bs. Prevents meaningful academic..... How the hell does agpl prevent academic use? Commercial use sure because agpl follows 4 freedoms and commercial often wants to take someone…",{"id":119,"by":120,"text":121},"35346590","alexb_",">GPL...prevents meaningful academic and commercial use WTF are you talking about?",{"value":123,"source":60,"evidence":124},"Code may be a derivative work of the original GPL code",[125],{"id":126,"by":127,"text":128},"35346110","querez","IANAL, but this seems very fishy to me: 1) I don't understand how this isn't a derivative work of the original code, as I very highly doubt you've done a clean room…",{"value":130,"source":60,"evidence":131},"Suspected coordinated promotion by Lightning AI staff",[132],{"id":133,"by":134,"text":135},"35347603","ipsum2","FYI, there's something fishy going on in this thread. Multiple people from the LightningAI team theaniketmaurya (developer advocate for Lightning AI) and rasbt (developer at…",[137,141],{"value":138,"source":60,"evidence":139},"Examples or demos of model capabilities",[140],{"id":88,"by":89,"text":90},{"value":142,"source":60,"evidence":143},"Clarity on how weights would be obtained or shared",[144,147],{"id":145,"by":82,"text":146},"35344897","I am curious though how would the model weights work out?",{"id":148,"by":149,"text":150},"35344925","rasbt","I guess that means time to fire up a few GPUs later today and get some weights! We should have a weight exchange platform for that maybe, haha.",[152,160],{"value":153,"source":60,"evidence":154},"Whether the weights are covered by the original Facebook license",[155,156],{"id":126,"by":127,"text":128},{"id":157,"by":158,"text":159},"35347122","rnosov","1) I've looked at both codebases and this one is definitely a derivative of the nanoGPT. You can compare all three implementations yourself as they are actually surprisingly…",{"value":161,"source":60,"evidence":162},"Whether the implementation is a clean-room rewrite or derivative",[163,164],{"id":126,"by":127,"text":128},{"id":70,"by":71,"text":72},[166,173,177,181],{"value":167,"source":60,"evidence":168},"llama.cpp (MIT licensed)",[169],{"id":170,"by":171,"text":172},"35347157","homarp","llama.cpp is also MIT https:\u002F\u002Fgithub.com\u002Fggerganov\u002Fllama.cpp previously discussed here https:\u002F\u002Fnews.ycombinator.com\u002Fitem?id=35100086 and one of the rust wrapper:…",{"value":174,"source":60,"evidence":175},"alpaca.cpp and alpaca-lora implementations",[176],{"id":63,"by":64,"text":65},{"value":178,"source":60,"evidence":179},"ChatLLaMA",[180],{"id":63,"by":64,"text":65},{"value":182,"source":60,"evidence":183},"Petals distributed inference",[184],{"id":185,"by":105,"text":186},"35346431","Right, so sort of like https:\u002F\u002Fgithub.com\u002Fbigscience-workshop\u002Fpetals but for the training phase. I suppose different training runs could be proposed via a RFC type of procedure.…",{"comments":39,"answered":39,"key":188},[],{"commenters":190,"max_depth":43,"maker_replies":39,"first_comment_mins":191,"lifespan_hours":192},28,6,136,[194,196,197,199,201,204,206,209,211,213,216,220,222,223,224,228,230,232,235,237,238,239,243,244,246,250,251,253,256,259,262,264,266,269,272,275,278,280,282],[111,39,39,195,112,39],"commenter\u002Fneutral\u002Fidea",[96,39,39,195,97,39],[198,39,39,195,149,39],"35346690",[200,39,39,195,71,39],"35346845",[202,39,39,195,203,39],"35347598","dTal",[63,39,39,205,64,39],"commenter\u002Fcritical\u002Fidea",[207,39,39,195,208,39],"35420961","7to2",[133,33,33,210,134,39],"maker\u002Fcritical\u002Fnone",[126,31,33,212,127,39],"project\u002Fcritical\u002Fidea",[214,31,39,195,215,39],"35346593","MacsHeadroom",[217,31,42,218,219,39],"35346885","project\u002Fneutral\u002Fidea","kristjansson",[70,31,42,221,71,39],"maker\u002Fneutral\u002Fexecution",[157,31,42,218,158,39],[115,42,42,218,116,39],[225,42,39,226,227,39],"35346391","other\u002Fneutral\u002Fnone","ftxbro",[119,229,33,212,120,39],4,[231,229,33,212,82,39],"35346697",[233,229,39,195,234,39],"35346866","n3t",[236,229,39,226,171,39],"35347237",[170,43,39,226,171,39],[104,191,33,212,105,39],[240,191,39,241,242,39],"35346212","other\u002Fneutral\u002Fidea","Ciantic",[185,191,39,241,105,39],[245,191,39,226,71,39],"35346863",[247,248,39,195,249,39],"35347088",7,"ficiek",[100,32,33,212,101,39],[252,32,33,212,149,39],"35346706",[254,32,42,255,82,39],"35346744","project\u002Fmixed\u002Fidea",[257,34,42,218,258,39],"35345738","AmuVarma",[260,34,42,218,261,39],"35345922","sp332",[263,34,42,218,258,39],"35346176",[265,34,33,212,258,39],"35346181",[267,34,33,212,268,39],"35346533","woodson",[270,34,39,195,271,39],"35347640","Tepix",[88,273,43,274,89,39],10,"project\u002Fpositive\u002Fidea",[81,276,43,277,82,39],11,"project\u002Fpositive\u002Fperformance",[145,276,42,279,82,39],"project\u002Fneutral\u002Fperformance",[148,281,39,226,149,39],12,[283,281,39,226,284,39],"35346491","A4ET8a8uTh0",[],[287,292,296,300,304,308],{"id":288,"name":289,"points":290,"category":291},"35316679","Apple Notes Liberator",563,"productivity",{"id":293,"name":294,"points":295,"category":20},"35334848","Regex.ai",294,{"id":297,"name":298,"points":299,"category":20},"35313171","GPT-4 Reverse Turing Test",288,{"id":301,"name":302,"points":303,"category":20},"35350662","Libraria",271,{"id":305,"name":306,"points":307,"category":291},"35327112","Open AI is not Open",241,{"id":309,"name":310,"points":311,"category":20},"35301657","llm-security",238,false,"v2",{"id":315,"name":5,"line":6,"ai":312,"cat":20,"year":13,"points":14,"comments":15,"rec":28,"s":316,"top":18,"kw":317,"maker":11,"domain":10,"tier":19,"attempts":39,"strip":318,"url":9,"mix":319,"praise":78,"objection":93,"sub":21},35344787,-0.34,[24,25,26,27],"000000011033330110001000011333311055300",[31,32,33,34],1793991062760]