tweet_id
stringlengths
10
19
full_text
stringlengths
16
359
expanded_url
stringlengths
43
52
embeddings
listlengths
1.54k
1.54k
1682487370925391872
4/ Different Interpretations 🤔
CI: "If we repeated this experiment 100 times, we'd expect the CI to contain the true parameter ~95 times."
CrI: "Given the data, there's a 95% probability that the true parameter lies within this interval."
https://twitter.com/i/web/status/1682487370925391872
[ -0.002111246809363365, 0.014097957871854305, 0.01615922339260578, 0.00663190009072423, 0.01043731439858675, -0.008038248866796494, 0.04205258935689926, 0.017179517075419426, -0.022267190739512444, 0.02066781371831894, 0.0019113245652988553, -0.015290596522390842, -0.010313224978744984, 0.0...
1515358690081226761
@WeatherStripApp if you like weather visualizations, https://t.co/zcAnhBmKof is another amazing tool that more people should know about https://t.co/YPIevpGtek
https://twitter.com/i/web/status/1515358690081226761
[ -0.030777739360928535, 0.016641942784190178, -0.03621241822838783, 0.0179090965539217, -0.022808754816651344, -0.021752793341875076, -0.015233995392918587, -0.02665245160460472, 0.003276998642832041, -0.0026223028544336557, -0.03956333547830582, -0.006057695485651493, -0.035339489579200745, ...
1682479383913586688
New favourite chart https://t.co/JHYztRW021
https://twitter.com/i/web/status/1682479383913586688
[ 0.02341766282916069, -0.037846729159355164, -0.008696398697793484, -0.01783805340528488, 0.060833048075437546, 0.0012635868042707443, -0.00780588760972023, 0.016363143920898438, -0.013169826939702034, 0.01978604681789875, -0.044358592480421066, -0.0026437053456902504, -0.02596396952867508, ...
1682504769204805633
I've already recreated Edward Tufte’s New York City’s Weather in 2003 in @matplotlib... https://t.co/GiTZrgoP4M So I'd say it's about time to recreate it in @bokeh as well! See the first part of my upcoming blog post. https://t.co/oVw0bnuI0K https://t.co/uyjiitcz49
https://twitter.com/i/web/status/1682504769204805633
[ -0.02889389544725418, -0.02317546121776104, 0.013735735788941383, -0.0030801238026469946, 0.011695490218698978, -0.02316109463572502, -0.007180369924753904, -0.016867943108081818, -0.032212886959314346, 0.00006740576645825058, 0.0015238979831337929, -0.02040245197713375, -0.04873599857091904...
1682135179840704514
This one falls into the extra spicy category so don’t do this if you haven’t been training your hamstrings and don’t try to do this one without warming up first. https://t.co/C3sNQGX1wT
https://twitter.com/i/web/status/1682135179840704514
[ 0.0023487042635679245, 0.03074914962053299, -0.017837228253483772, -0.009789385832846165, -0.03942966088652611, 0.008265534415841103, -0.024082297459244728, -0.020218245685100555, 0.027320483699440956, -0.011803047731518745, 0.020177427679300308, 0.0004583460104186088, -0.02678985521197319, ...
1682109194172878848
New blog post: perspectives on diffusion, or how diffusion models are autoencoders, deep latent variable models, score function predictors, reverse SDE solvers, flow-based models, RNNs, and autoregressive models, all at once! https://t.co/sBG8Waa9Ql
https://twitter.com/i/web/status/1682109194172878848
[ -0.0015671799192205071, -0.014215262606739998, 0.028113456442952156, -0.012451565824449062, 0.027822809293866158, -0.010958699509501457, -0.027769966050982475, -0.007860670797526836, -0.06811700016260147, 0.006480098702013493, 0.017954034730792046, -0.021679596975445747, -0.00495420396327972...
1682010345177579523
Lars maakte een supergedetailleerde metro-stijl kaart van het Belgische treinnetwerk: https://t.co/bhxDtUhjY2 (via https://t.co/ZVZk68MUcH)
https://twitter.com/i/web/status/1682010345177579523
[ -0.04212597757577896, 0.004342728294432163, -0.014901459217071533, -0.02152298577129841, -0.064738430082798, -0.0012937396531924605, -0.007928883656859398, 0.029633445665240288, -0.04788803681731224, 0.008400985039770603, -0.03508077189326286, -0.03171553462743759, 0.011288067325949669, -0...
1681700743693172738
Thanks to everyone who came to my @EuroPython keynote on LLMs from prototype to production ✨ Here are my slides and a walkthrough of the talk as a Twitter thread 🧵 https://t.co/gNNtipUdNi https://t.co/mvqyCdsr8w
https://twitter.com/i/web/status/1681700743693172738
[ 0.0068818978033959866, 0.013902823440730572, 0.0024709110148251057, -0.0225225742906332, -0.015343662351369858, -0.02499980479478836, -0.01980520412325859, 0.05535851791501045, 0.001060090260580182, 0.027603425085544586, -0.0019163779215887189, -0.012898028828203678, 0.0008689264650456607, ...
1682064328550236163
Chiin is giving a lightning talk @EuroPython #EuroPython2023 about environmental conscious science education https://t.co/IiFWtob8wY
https://twitter.com/i/web/status/1682064328550236163
[ 0.0005860139499418437, -0.023345159366726875, 0.03186280280351639, 0.037232328206300735, 0.02801964432001114, -0.001361971371807158, -0.014323270879685879, 0.05235966295003891, -0.008606228046119213, -0.00620084535330534, -0.016176709905266762, 0.005488770082592964, -0.06083642318844795, 0...
1682034962424537090
As the number of #Python function arguments increases, it becomes challenging for developers to keep track of the purpose of numerous arguments and use the function. To improve code readability, you can bundle multiple related arguments into a data structure with a dataclass. https://t.co/9nEK1LlqcW
https://twitter.com/i/web/status/1682034962424537090
[ 0.000552720797713846, -0.011490494012832642, 0.013222605921328068, -0.009130185469985008, 0.028396887704730034, 0.0125639159232378, -0.031690340489149094, -0.005830632522702217, 0.004421766381710768, 0.03966781869530678, 0.013673931360244751, -0.04183905944228172, 0.012746885418891907, -0....
1681676861443637251
Check out the latest release of #Leafmap v0.23.0! 🌿🗺️ New features include interactive extraction and visualization of #AWS Open #Geospatial Data. 🌐🔍 Now you can easily extract sample images for your region of interest without downloading large amounts of data. 🙌 Perfect for… https://t.co/pDX5N6WKE1
https://twitter.com/i/web/status/1681676861443637251
[ -0.0034635341726243496, 0.021488282829523087, 0.03419269993901253, -0.022534528747200966, -0.01417605858296156, -0.023684250190854073, 0.044333238154649734, 0.01637202501296997, 0.01162942685186863, 0.02487996034324169, 0.03814774006605148, -0.015176315791904926, -0.045046065002679825, -0....
1681663852256145410
To get estimated prediction intervals for predictions made by a scikit-learn model, use MAPIE. In the following code, we use MapieRegressor to estimate prediction intervals for a scikit-learn regressor. https://t.co/SdHMlXozom #Python https://t.co/UsvBKYgUeP
https://twitter.com/i/web/status/1681663852256145410
[ 0.021516017615795135, -0.00039774327888153493, 0.04195455089211464, -0.011313586495816708, -0.013603241182863712, -0.008266326040029526, 0.0019192692125216126, 0.011964567005634308, -0.01692548394203186, 0.0016232415800914168, -0.0011630658991634846, -0.05162946134805679, -0.0103932349011302...
1681641346396835841
LLAMA-v2 training successfully on Google Colab's free version! "pip install autotrain-advanced" 💥 Yes, you can also use your local machine! https://t.co/VOvocAQ46c
https://twitter.com/i/web/status/1681641346396835841
[ 0.008213371969759464, -0.023667404428124428, 0.043858371675014496, -0.036168765276670456, 0.03817174583673477, -0.03605365380644798, 0.018475770950317383, 0.0208931602537632, -0.033935558050870895, 0.002617400838062167, -0.030827485024929047, 0.014055399224162102, -0.014757594093680382, 0....
1681602528927170563
This Post is from a suspended account. {learnmore}
https://twitter.com/i/web/status/1681602528927170563
[ -0.00972004234790802, 0.0030913425143808126, -0.014003320597112179, 0.06164229288697243, -0.007343861274421215, -0.031805455684661865, -0.04269436374306679, 0.026745496317744255, 0.0015562449116259813, -0.0059481430798769, 0.007074714172631502, -0.012273091822862625, -0.026084164157509804, ...
1681280868315193344
LMQL now supports @ggerganov's excellent llama.cpp as (CPU-based) inference backend. Seamlessly switch your query code from OpenAI to🤗 to llama.cpp. This is possible due to our new token streaming protocol LMTP. LMTP: https://t.co/zZ16urbUto Docs: https://t.co/wWsQt9LJkq https://t.co/iKcYtbRsru
https://twitter.com/i/web/status/1681280868315193344
[ 0.013197972439229488, 0.013021811842918396, 0.027988439425826073, -0.023084120824933052, 0.03599318861961365, -0.022703614085912704, -0.005313011817634106, 0.015854477882385254, -0.019222674891352654, -0.039460036903619766, -0.0003741216496564448, -0.012690629810094833, 0.010055262595415115,...
1681359703350333441
Excited for Spy, a static python to WASM compiler. Amazing work from @antocuni EuroPython Slides: https://t.co/ZCbjOfyYb2 Repo: https://t.co/31l965C6dm
https://twitter.com/i/web/status/1681359703350333441
[ -0.027021214365959167, -0.013696874491870403, -0.015298775397241116, -0.02640032209455967, -0.034769944846630096, 0.008021922782063484, 0.018216967582702637, -0.004029588308185339, 0.039538394659757614, -0.022612880915403366, 0.007276852615177631, -0.04212130606174469, -0.0011835751356557012...
1681378525830414338
We're very excited to be one of the launch partners for Meta's Llama 2 🦙! We got to test Llama 2 in advance and were very impressed. The new version also has a much more permissive license. We've set everything up so you can run it on Databricks today. https://t.co/gf0OQZLBvZ
https://twitter.com/i/web/status/1681378525830414338
[ 0.0259220190346241, -0.016503896564245224, 0.01820245385169983, -0.052579235285520554, 0.02388121373951435, 0.010273738764226437, 0.03970061615109444, 0.04053721949458122, -0.023576995357871056, -0.016795439645648003, -0.01969819888472557, -0.018151750788092613, -0.04497375339269638, -0.01...
1681574687489368064
Hiep, hiep, hoera voor onze favoriete entomoloog, Peter Berx (@Entomobiel)! 🥳 We zetten hem met plezier even in de bloemetjes (en bijtjes 😁) met deze video van zijn passage bij Vanavond Live met @xanderycke. Gefeliciteerd, Peter! https://t.co/l7AfxQjYSo
https://twitter.com/i/web/status/1681574687489368064
[ 0.03288073465228081, 0.006127054803073406, -0.04037311673164368, 0.013145523145794868, 0.016745027154684067, 0.008034002035856247, 0.019125889986753464, 0.06391093879938126, -0.002129800384864211, -0.022928500548005104, -0.030466018244624138, -0.019419265910983086, 0.020908715203404427, -0...
1681404748904431616
Llama 2 has been released today, and of course I had to test it on my Mac 😇 ! This is the 7B chat model, converted to Core ML and running locally at ~6.5 tokens per second: https://t.co/jtP22mfCiY
https://twitter.com/i/web/status/1681404748904431616
[ -0.0018251417204737663, -0.013125996105372906, 0.03832627087831497, -0.04275168105959892, -0.0034778425469994545, 0.01068792026489973, 0.015488948673009872, 0.05796745792031288, -0.0102645019069314, -0.017920194193720818, -0.00973181240260601, -0.004538097884505987, -0.02410757914185524, 0...
1681551459970580480
How to train LLAMA-v2 on custom dataset in just 50 lines of code: https://t.co/Uod2KxQKJS
https://twitter.com/i/web/status/1681551459970580480
[ -0.016507893800735474, -0.019957304000854492, 0.030655663460493088, -0.05918462201952934, 0.003653652034699917, -0.015859507024288177, 0.015392670407891273, 0.020333366468548775, -0.007417529355734587, 0.06068887561559677, -0.02401619590818882, -0.04525730386376381, 0.016287442296743393, 0...
1681503574814081025
@simonw Yes: https://t.co/oBttZwKtpV Getting 20-25 token/s with https://t.co/cS1t8gnUSi on M2 Max
https://twitter.com/i/web/status/1681503574814081025
[ 0.007453629747033119, 0.01995386742055416, 0.0109226880595088, -0.04806292802095413, 0.0024322166573256254, -0.014356860890984535, -0.017426686361432076, 0.06561364978551865, 0.030155610293149948, -0.04545822739601135, 0.021070167422294617, -0.017287150025367737, -0.057954587042331696, 0.0...
1681463685737246721
These lines stick out from one of my favorite McKinsey studies: 1) "The role of industry in a company’s position is so substantial that you’d rather be an average company in a great industry than a great company in an average industry." 2) "In some cases, you’d rather be in… https://t.co/pDohzsYtI7
https://twitter.com/i/web/status/1681463685737246721
[ -0.01994599588215351, -0.02878601662814617, 0.034688688814640045, 0.039584267884492874, 0.06495736539363861, 0.002173288259655237, -0.012826422229409218, 0.04210199788212776, -0.03600350022315979, 0.010658378712832928, 0.01342787966132164, -0.03508033603429794, 0.030688298866152763, -0.001...
1681208389982928896
#van9tot5 https://t.co/8TY7T9ZTlH
https://twitter.com/i/web/status/1681208389982928896
[ -0.0023756392765790224, 0.004728653468191624, -0.03991073742508888, -0.0062520792707800865, 0.018416861072182655, -0.040815744549036026, -0.0022285757586359978, 0.04295758903026581, -0.0008097913232631981, -0.02565690316259861, 0.00279043335467577, -0.02371114119887352, -0.003235394367948174...
1681395826684305425
@pydatalondon @HelloFresh @pymc_labs @pymc_devs Notebook can be found here: https://t.co/wxJvXam8iG
https://twitter.com/i/web/status/1681395826684305425
[ -0.031688299030065536, -0.04095590114593506, 0.030249228700995445, -0.029990196228027344, 0.03186098486185074, -0.014980707317590714, -0.0012375992955639958, 0.046280454844236374, 0.010361296124756336, -0.010382882319390774, -0.010958509519696236, -0.020449168980121613, -0.005641150288283825...
1681053664365121536
Prank https://t.co/azXqDOykUu
https://twitter.com/i/web/status/1681053664365121536
[ -0.011982927098870277, 0.04002012312412262, -0.04782295972108841, 0.020839283242821693, -0.03148322179913521, -0.049916405230760574, -0.03588761016726494, 0.02919946424663067, -0.02181803621351719, -0.03923168405890465, -0.011846989393234253, -0.03262510150671005, -0.0025709259789437056, 0...
1852289411855057016
You can watch the most information-dense recording on retrieval in the RAG context this weekend. This includes practical thinking around hybrid search, ColBERT, ColPali, and binary vectors, as well as how Ravenpack uses Vespa's SPANN implementation to scale to B of vectors.…
https://twitter.com/i/web/status/1852289411855057016
[ 0.04679306223988533, 0.03376208245754242, 0.016020771116018295, -0.016443856060504913, 0.007277040742337704, 0.00435071112588048, 0.049472592771053314, 0.02559656649827957, -0.026611968874931335, -0.0066036321222782135, 0.04560842737555504, -0.024947838857769966, 0.02865687385201454, -0.02...
1849533633977369017
I use the VLM to describe my screenshot using a structured output schema. https://t.co/oNUqZqxyEi
https://twitter.com/i/web/status/1849533633977369017
[ -0.03236694633960724, 0.0032087310682982206, 0.017585191875696182, -0.03845521807670593, -0.01121374499052763, 0.009089929983019829, 0.0033149218652397394, -0.019624054431915283, -0.01795331947505474, 0.0009627962717786431, 0.0016202940605580807, 0.0031273181084543467, 0.0008818258065730333,...
1852114022126030913
https://t.co/PLbZbvgdcM
https://twitter.com/i/web/status/1852114022126030913
[ 0.0283440500497818, 0.034627076238393784, -0.02008877508342266, 0.018651850521564484, 0.02000425010919571, -0.061928652226924896, 0.04513635113835335, 0.067507304251194, -0.020469138398766518, -0.021779276430606842, -0.05798415094614029, 0.03324650228023529, -0.013298599980771542, -0.02935...
1851973388106977779
The Dutch DPA calls for input on the AI Act prohibition of AI systems on emotion recognition in the areas of workplace or education institutions. See https://t.co/0vykTXa8hY. https://t.co/RMot2rA9yG
https://twitter.com/i/web/status/1851973388106977779
[ 0.020626721903681755, -0.009572514332830906, 0.04538658633828163, 0.0408635251224041, 0.00014814895985182375, 0.02026279829442501, -0.0072135040536522865, 0.051833249628543854, 0.03093358501791954, 0.02669646218419075, 0.010527816601097584, -0.08032333850860596, -0.010527816601097584, -0.0...
1852138612932186394
https://t.co/Hp0eZ3fjff
https://twitter.com/i/web/status/1852138612932186394
[ -0.015074440278112888, -0.007372901309281588, 0.005804731044918299, 0.019975420087575912, -0.011859511956572533, -0.061097919940948486, 0.03880775347352028, 0.03975079953670502, 0.03983652964234352, -0.02067555859684944, -0.019746802747249603, 0.012595373205840588, -0.05341067165136337, 0....
1852138610985963973
"An open-source, lightweight note-taking solution. The pain-less way to create your meaningful notes." https://t.co/sN8FOL9F7k
https://twitter.com/i/web/status/1852138610985963973
[ -0.010301911272108555, -0.011041436344385147, 0.010179729200899601, -0.008719969540834427, 0.004700809717178345, -0.04185071215033531, 0.010173298418521881, 0.012629808858036995, -0.01683545857667923, -0.037683647125959396, -0.032384783029556274, 0.026391413062810898, -0.04056457802653313, ...
1852135739947192604
I’ve literally been telling image gen people for a few months now that, in my opinion, we should stop one-shotting single images, and instead create objects and layers, and have a « make consistent » model to blend them. Looks like Blockade Labs is getting closer to what i want: https://t.co/vxgQiTZLBr
https://twitter.com/i/web/status/1852135739947192604
[ 0.007439234759658575, -0.01128228660672903, -0.05798108130693436, 0.04168035089969635, -0.0031669253949075937, -0.029255831614136696, 0.026897680014371872, 0.005954335443675518, -0.018909437581896782, 0.015932269394397736, -0.0022347180638462305, -0.02918213978409767, -0.019616883248090744, ...
1851999764033790196
https://t.co/fA5mfRyhQx
https://twitter.com/i/web/status/1851999764033790196
[ -0.02517288364470005, 0.004019947722554207, -0.007832222618162632, 0.017889508977532387, 0.011384907178580761, -0.030557535588741302, -0.024727869778871536, 0.02836214005947113, 0.02474270388484001, -0.003934653475880623, -0.02011457271873951, -0.04212786629796028, -0.037025053054094315, -...
1851732285952844241
https://t.co/9QOtTSqyDT
https://twitter.com/i/web/status/1851732285952844241
[ -0.017373880371451378, -0.010851867496967316, 0.0020270661916583776, -0.014705166220664978, 0.000026766401788336225, -0.02188074216246605, -0.017537271603941917, 0.05941976234316826, -0.041855260729789734, -0.05647873133420944, 0.01198198739439249, 0.0006229271530173719, -0.00619523180648684...
1851919137481609347
Don’t sleep on Vision Language Models (VLMs). With the releases of Llama 3.2 and ColQwen2, multimodal models are gaining more and more traction. VLMs are multimodal models that can handle image and text modalities: Input: Image and text Output: Text They can be used for many… https://t.co/DPW1ParoXa
https://twitter.com/i/web/status/1851919137481609347
[ -0.0036045031156390905, 0.03441397845745087, -0.013247638009488583, -0.015513681806623936, -0.003436417318880558, -0.0027562931645661592, -0.006816806271672249, 0.011392470449209213, -0.023544440045952797, 0.01223912462592125, -0.035459842532873154, -0.016198474913835526, -0.0066424952819943...
1851721189133291888
🇪🇺 EU Policy Brief: "Harmonised Standards for the European AI Act" (7 pages) is the document every legal team should be reading, clarifying the role of standards in the EU AI Act. Download it below👇 https://t.co/Stq15tQjE9
https://twitter.com/i/web/status/1851721189133291888
[ -0.0020086595322936773, 0.030340099707245827, 0.0044114599004387856, -0.026042502373456955, -0.0045720357447862625, 0.02224707044661045, -0.014983197674155235, -0.007345621008425951, 0.0003288158040959388, 0.006072691176086664, 0.012390625663101673, -0.04867495596408844, -0.03248889744281769...
1851629939629375765
Want to learn more about multiple imputation for #missingdata? This book by Stef van Buuren is the perfect guide! It addresses the often-overlooked topic of multiple imputation compared to univariate methods. Plus, it's🆓! Check it out!👇 https://t.co/CUV5yyIXqj https://t.co/KLM10Vv4xF
https://twitter.com/i/web/status/1851629939629375765
[ -0.045420143753290176, 0.02215499058365822, 0.029491718858480453, -0.0010128727881237864, -0.03410130739212036, -0.04749566689133644, -0.007004886865615845, 0.03868676349520683, -0.046313099563121796, 0.011240399442613125, 0.0259681586176157, -0.01694205217063427, -0.017014453187584877, -0...
1851663863026266263
With the help of @hugovk we now have GitHub action powered by uv for pre-commit, https://t.co/sJNa7qzwoN (caching included), great to use whenever https://t.co/A7oZMSNhLX is not an option.
https://twitter.com/i/web/status/1851663863026266263
[ -0.04369785264134407, 0.029837273061275482, 0.004176028072834015, -0.0058160200715065, 0.0199576448649168, -0.016320565715432167, -0.004767879843711853, -0.007703994866460562, -0.005683762487024069, -0.024004722014069557, 0.006265695206820965, 0.00123578030616045, -0.00313119450584054, -0....
1851695007717290257
Eugene created a game to help people LOOK AT THE DATA. It's simple but very effective in my experiments! This is the tool we need. https://t.co/f3qnIljdoA https://t.co/7rZho7AyB5 https://t.co/iHQnZSNbbB
https://twitter.com/i/web/status/1851695007717290257
[ -0.02722816728055477, 0.004965042695403099, 0.009229887276887894, -0.0044995700009167194, 0.0023731153924018145, -0.05041428282856941, -0.01049501821398735, 0.015308882109820843, 0.006631992291659117, -0.008322812616825104, 0.03653762489557266, -0.035964734852313995, 0.013789133168756962, ...
1851570743949090964
My PyData Amsterdam talk is out! The title is "Run a benchmark they said, it will be fun they said". The talk is, as the name implies, about a benchmark. Enjoy! https://t.co/Aw1aULe6hP
https://twitter.com/i/web/status/1851570743949090964
[ 0.010909699834883213, -0.02959471382200718, 0.03490593284368515, -0.02331317961215973, -0.017759382724761963, 0.013622760772705078, -0.023364247754216194, 0.03345045447349548, 0.027756216004490852, -0.011228883638978004, 0.021423611789941788, -0.028215840458869934, -0.014056851156055927, -...
1846611547143262585
Now all my projects (@FastAPI, Typer, SQLModel, Asyncer, etc) use uv to install packages in development and CI. 🚀 Much simpler, faster, clearer. ✨
https://twitter.com/i/web/status/1846611547143262585
[ -0.07118725776672363, -0.002790494356304407, 0.013933834619820118, -0.002690531313419342, 0.028084538877010345, -0.029060449451208115, 0.008112256415188313, -0.026485128328204155, 0.008586658164858818, 0.01163637824356556, 0.011012880131602287, 0.01519438624382019, 0.014516670256853104, -0...
1851663992671842795
@trickylabyrinth This is an all-timer https://t.co/Ix36Ip6ddF
https://twitter.com/i/web/status/1851663992671842795
[ -0.001144171692430973, 0.03015228733420372, -0.022306539118289948, 0.008268803358078003, 0.02589096873998642, 0.0028171620797365904, -0.012099375016987324, 0.023568011820316315, -0.004265164025127888, 0.015860719606280327, 0.0027825485449284315, -0.033813636749982834, 0.02256806381046772, ...
1851645681150382103
The most common mistakes I see teams make with LLM judges: • Too many metrics • Complex scoring systems • Ignoring domain experts • Unvalidated measurements That's why I wrote this guide, w/ detailed examples to help people avoid these issues (1/4) https://t.co/qWoti5QDvO
https://twitter.com/i/web/status/1851645681150382103
[ 0.016372309997677803, 0.03242586925625801, 0.033092353492975235, 0.012981938198208809, 0.022414127364754677, -0.0008557430119253695, -0.04343733564019203, -0.020385699346661568, -0.01638679951429367, 0.046711795032024384, -0.002086383057758212, -0.049899324774742126, -0.020950762555003166, ...
1851654159692616020
Evaluating LLM output is hard. It's the bottleneck to scaling AI products for many teams A key mistake is defining eval criteria w/o actually LOOKING AT THE DATA. This leads to irrelevant / unrealistic criteria + wasted effort Thus, I built AlignEval https://t.co/vWdlRoctwQ
https://twitter.com/i/web/status/1851654159692616020
[ 0.0021826087031513453, 0.01949087344110012, 0.05042358487844467, -0.01050491351634264, 0.01960444077849388, -0.023607663810253143, 0.010817221365869045, 0.003346666693687439, 0.005518628749996424, 0.0530640073120594, 0.010029353201389313, -0.05647100880742073, -0.023465706035494804, -0.001...
1851326301237374987
https://t.co/mdWRUhuVBe
https://twitter.com/i/web/status/1851326301237374987
[ -0.029276302084326744, 0.0005680514732375741, -0.0061358134262263775, 0.02011202462017536, -0.014391209930181503, -0.05004683881998062, 0.03577909991145134, 0.06387557089328766, -0.008773287758231163, 0.016449056565761566, -0.025270359590649605, 0.014377490617334843, 0.003858462907373905, ...
1851252352197480527
ColPali is changing the game for PDF retrieval by eliminating the need for OCR and chunking methods 🚀 Inspired by ColBERT’s success with text, ColPali splits an image of a document into patches, which are then processed through a vision LLM called PaliGemma. The embeddings for… https://t.co/E2zwupDGQ8
https://twitter.com/i/web/status/1851252352197480527
[ -0.016697222366929054, -0.009087014012038708, 0.02712799608707428, 0.030201349407434464, 0.009512759745121002, 0.011947494000196457, 0.023722030222415924, 0.038609832525253296, -0.025491535663604736, -0.045129064470529556, 0.02132720872759819, -0.04409131035208702, -0.018932387232780457, 0...
1851319613889359998
Using VLMs for OCR https://t.co/2UzEjX2laK
https://twitter.com/i/web/status/1851319613889359998
[ -0.0439094640314579, 0.050458863377571106, 0.02911635860800743, -0.032405298203229904, -0.027222728356719017, 0.000768842757679522, 0.00146560650318861, 0.00985969603061676, -0.03374365344643593, 0.024531777948141098, -0.003207070752978325, -0.01422359049320221, -0.026738641783595085, 0.00...
1851322576728977533
Trending new agent framework https://t.co/sBYPfEhvef
https://twitter.com/i/web/status/1851322576728977533
[ -0.07226195931434631, 0.003704268019646406, 0.0268760547041893, 0.011568655259907246, 0.00008900347165763378, 0.021160678938031197, 0.020164702087640762, 0.021834880113601685, 0.021712297573685646, 0.011346476152539253, 0.0027006298769265413, -0.0592682883143425, -0.032422881573438644, -0....
1851335124853408002
Releasing Florence-2-DocLayNet-Fixed: Florence-2 modelf finetuned on the DocLayNet dataset. To prevent the model from generating hallucinated class names, we re-mapped all class names to single tokens. This simple change brought 7% improvement of mAP50-95 score on the DocLayNet… https://t.co/qWiVH1Iqaa
https://twitter.com/i/web/status/1851335124853408002
[ 0.025069663301110268, -0.005304679274559021, 0.037354182451963425, -0.0579695850610733, -0.022168615832924843, -0.00924869067966938, -0.022194288671016693, 0.0009474933613091707, -0.010230681858956814, 0.017573153600096703, -0.017996758222579956, -0.011764641851186752, -0.046545110642910004,...
1851321941841330526
https://t.co/cX5mAKkMUn
https://twitter.com/i/web/status/1851321941841330526
[ 0.003191232681274414, 0.023762844502925873, -0.05999376252293587, -0.011485611088573933, -0.001952553866431117, -0.028512587770819664, -0.028795309364795685, 0.02331048808991909, 0.03152358531951904, -0.012475140392780304, 0.012086396105587482, -0.009068331681191921, -0.0037566781975328922, ...
1681040273605029889
🦜🔗+Streamlit: Chat With Your Documents Shout out to @meShashank93 for a FANTASTIC end-to-end example of using @LangChainAI and @streamlit to chat with your documents! Code: https://t.co/45rZX6B4gG https://t.co/9CXI7yHbsu
https://twitter.com/i/web/status/1681040273605029889
[ 0.009925510734319687, -0.006376075558364391, 0.0045497301034629345, -0.006963345687836409, 0.041483212262392044, -0.03833389654755592, 0.007014973554760218, 0.04295461252331734, -0.02031567320227623, -0.0203285813331604, 0.0013858929742127657, -0.031906191259622574, -0.032061077654361725, ...
1681245990911725568
In my latest article from the Writing Faster Python series, I've run a series of benchmarks for different tasks to see how much faster (or slower) it is to use pathlib instead of other functions. https://t.co/uXEC9YgrkT
https://twitter.com/i/web/status/1681245990911725568
[ -0.002201144816353917, -0.0038940978702157736, 0.043324898928403854, -0.04680264741182327, 0.05177435651421547, 0.0108067337423563, 0.017805088311433792, 0.04469640552997589, 0.007843300700187683, 0.019494980573654175, 0.013690570369362831, -0.006575881969183683, -0.02062157541513443, 0.00...
1680958393807167488
I didn't get a chance to present my lightning talk at #SciPy2023, so thought I'd record and share it here Declarative visualization in Python using @vega_vis Vega-Altair https://t.co/LwDSYeZd48
https://twitter.com/i/web/status/1680958393807167488
[ -0.008775964379310608, -0.03832848370075226, -0.004155310802161694, -0.03423871472477913, -0.011010918766260147, 0.023909423500299454, -0.02695053443312645, -0.003150892211124301, -0.006803174503147602, 0.00022427372459787875, -0.008113997988402843, 0.006967027671635151, 0.006724525243043899...
1680584420648968192
Our own @stefdegreef has fully covered the entire Belgian capital #Brussels by bike to create street-level imagery in 360° using one of our GoPro Hero Max's. All footage is available on @mapillary .com under the CC-BY-SA license. Massive applause👏👏👏 https://t.co/JrfLcNmj6j https://t.co/yOa0fYPCVm
https://twitter.com/i/web/status/1680584420648968192
[ 0.023402879014611244, -0.0017493192572146654, 0.03409534692764282, -0.02207942306995392, -0.07254105806350708, -0.036820877343416214, -0.005804857239127159, 0.035772595554590225, -0.015082146041095257, -0.02206631936132908, 0.026075996458530426, -0.026914620772004128, -0.04153814539313316, ...
1680676258156257283
Object Detection and Image Classification with YOLO #KDnuggets https://t.co/eb96YIuJBj
https://twitter.com/i/web/status/1680676258156257283
[ -0.03439049795269966, -0.012702591717243195, -0.04743517190217972, -0.0139226783066988, -0.017833799123764038, 0.01708122342824936, -0.0023503785487264395, 0.04098125174641609, -0.03717275336384773, 0.026431424543261528, 0.008415182121098042, -0.010131285525858402, 0.0020425058901309967, -...
1679931131796307975
GPT4All now supports Text Embeddings ⚡ - Generate text embeddings of arbitrary length documents for free on CPU at 8,000 tok/second. - No external dependencies except C. https://t.co/W3jvubmTyM https://t.co/4PnLwC15mQ
https://twitter.com/i/web/status/1679931131796307975
[ -0.03329271078109741, 0.011753794737160206, 0.059036899358034134, 0.019453752785921097, -0.013617627322673798, -0.024462804198265076, -0.0010549584403634071, 0.04158676415681839, -0.008002831600606441, -0.04799368977546692, 0.011363554745912552, 0.02651301957666874, 0.01062384620308876, -0...
1680665487598837760
The essay I just published called "How to Do Great Work" (https://t.co/zzxQTx7JJ3) grew out of a single paragraph in another essay I was writing. It seemed such an important topic that I cut it out and made it into its own essay.
https://twitter.com/i/web/status/1680665487598837760
[ 0.01273350790143013, 0.01264796033501625, 0.00827842578291893, 0.035719629377126694, 0.031744930893182755, -0.025743402540683746, -0.042537156492471695, 0.052408091723918915, -0.03261357545852661, -0.005399402230978012, 0.04135264456272125, 0.0021222513169050217, 0.010450031608343124, 0.01...
1680505967606104064
Coding tutorial on how to create a falcon (or any LLM) based PDF chatbot that runs locally without using langchain. Check it out here: https://t.co/LPb6j029Pn https://t.co/ipHDeRs3Hu
https://twitter.com/i/web/status/1680505967606104064
[ -0.03654760494828224, -0.005460841115564108, 0.022901011630892754, -0.026059266179800034, 0.010951061733067036, 0.022577842697501183, 0.009739173576235771, 0.045978300273418427, -0.033022113144397736, 0.05388127639889717, 0.01477768924087286, -0.026191471144557, 0.01015782542526722, -0.016...
1680190180102217729
I finally found the time to wrap up the calmcode course on embeddings. https://t.co/xEW6bPHJzs The course starts by training letter embeddings from scratch and will use that experiment as a vehicle to explain how you might also get to multi-modal representations too.
https://twitter.com/i/web/status/1680190180102217729
[ 0.02162357047200203, 0.013630767352879047, 0.006719824858009815, -0.024094445630908012, -0.02103656716644764, -0.005969006568193436, 0.020244793966412544, 0.047233302146196365, 0.0008020105888135731, 0.035820864140987396, 0.049990855157375336, -0.010490980930626392, -0.011596731841564178, ...
1679845079261822976
Jupyter-scatter is a brand new library by @flekschas for producing, well, 2d scatter plots in a Jupyter notebook, but with amazing quality and easy to use: https://t.co/1GHS4BMFjs #SciPy2023 🤩 https://t.co/Ci391Ufao0
https://twitter.com/i/web/status/1679845079261822976
[ -0.0597323402762413, -0.04964606463909149, 0.019215457141399384, -0.012337894178926945, 0.012276542373001575, -0.0367375984787941, 0.004763985518366098, 0.0032669957727193832, -0.02198856882750988, -0.0005728746182285249, -0.031191373243927956, -0.0037486092187464237, -0.0368112213909626, ...
1679778530966601730
Showing the trend gives us more to work with, shows us patterns, gives us real information. 💟 As always, Edward Tufte said it best: "For non-data-ink, less is more. For data-ink, less is a bore." (6/6) #informationdesign #datavisualization #powerfulcharts https://t.co/ZBeTfBPsdG
https://twitter.com/i/web/status/1679778530966601730
[ 0.00897629838436842, 0.011382832191884518, 0.017352748662233353, 0.009133401326835155, 0.08335032314062119, 0.004691669251769781, 0.006526918616145849, 0.006362674292176962, -0.023365512490272522, 0.02212296985089779, 0.014981919899582863, 0.007048215251415968, -0.02497938834130764, -0.025...
1679595929408610304
The CEO of DeepMind, Demis Hassabis, went on @ezraklein to talk about the future of AI and science. These are his book recommendations if you want to expand your mind and see what an AGI future looks like: Permutation City, by Greg Egan — a wild story about how interesting and… https://t.co/iiBiZzbvlU
https://twitter.com/i/web/status/1679595929408610304
[ 0.0133902532979846, -0.03169068321585655, 0.005079277791082859, 0.04146093502640724, 0.014555167406797409, -0.06784061342477798, -0.010903850197792053, 0.07721003144979477, -0.007647099904716015, -0.03164058178663254, 0.013966447673738003, -0.050103846937417984, 0.0027212651912122965, -0.0...
1679521720179339264
The magic of @duckdb: A spatial query (st_buffer) on a @GeoParquet file in a GitHub repository in seconds... Goodbye to the database. https://t.co/82xZmZTYS9
https://twitter.com/i/web/status/1679521720179339264
[ -0.010318771935999393, -0.005983318202197552, 0.06837330758571625, -0.004999177064746618, 0.020101334899663925, -0.028196796774864197, 0.01301943976432085, 0.039888788014650345, -0.024861833080649376, -0.04195515811443329, 0.015890125185251236, -0.009298665449023247, 0.010907295159995556, ...
1679346323097636864
Tutorial on using @duckdb with the Google Open Buildings @GeoParquet files on https://t.co/KKuNdQ9Qy6 https://t.co/IH1qGVc1MQ
https://twitter.com/i/web/status/1679346323097636864
[ -0.014026468619704247, 0.007923544384539127, 0.06872200220823288, -0.026091283187270164, -0.007545317057520151, -0.03251473978161812, 0.0000385890161851421, 0.01451367698609829, -0.03051462024450302, -0.03841252252459526, 0.023360352963209152, 0.0011891727335751057, 0.014782923273742199, -...
1679358632431853568
this is wild — kNN using a gzip-based distance metric outperforms BERT and other neural methods for OOD sentence classification intuition: 2 texts similar if cat-ing one to the other barely increases gzip size no training, no tuning, no params — this is the entire algorithm: https://t.co/7mLIRlX48N https://t.co/IWe402RGgn
https://twitter.com/i/web/status/1679358632431853568
[ -0.023277517408132553, 0.005908908322453499, 0.02701215259730816, 0.014439734630286694, 0.01318632997572422, -0.019798679277300835, -0.006091163493692875, 0.024684401229023933, -0.031335119158029556, 0.034609321504831314, 0.05233604460954666, -0.003830558620393276, -0.027498167008161545, 0...
1679145605766189057
New in @scikit_learn -- experimental support for building models on GPUs with @PyTorch - @thomasjpfan at #SciPy2023 https://t.co/3zdcykW95H
https://twitter.com/i/web/status/1679145605766189057
[ 0.00812933873385191, -0.04056665673851967, 0.037365227937698364, -0.013388825580477715, 0.03663347288966179, 0.006837334018200636, 0.00044912879820913076, 0.07098021358251572, -0.02283303625881672, -0.00009807943570194766, -0.01691039651632309, -0.011868148110806942, -0.02119802124798298, ...
1679143673693609985
This is a wonderful overview of plotting with various Python libraries! 📈🐍 https://t.co/nUY8CvMlUw
https://twitter.com/i/web/status/1679143673693609985
[ -0.02929910458624363, -0.00881852675229311, 0.008806528523564339, -0.03213063254952431, -0.010348270647227764, -0.008824525400996208, 0.016845185309648514, 0.013749702833592892, -0.021152464672923088, 0.010204294696450233, 0.01832093857228756, -0.0026185624301433563, -0.014349602162837982, ...
1679199371169251328
ICYMI - there are now Azure Developer CLI Learn modules available! ✨Learn how the Azure Developer CLI can accelerate the time it takes for you to get your application from local development environment to Azure. ✨ https://t.co/PJNhwuB3m7
https://twitter.com/i/web/status/1679199371169251328
[ -0.038651008158922195, 0.0005298560718074441, 0.02370108850300312, 0.003859973046928644, 0.033363841474056244, -0.004045137669891119, 0.02711951546370983, 0.01614636741578579, -0.024453142657876015, 0.0391295850276947, -0.009224053472280502, -0.035825107246637344, -0.00013175184722058475, ...
1679013217182396419
“Tour de France” by Jean-Jacques Sempé The New Yorker magazine July 12, 1999 https://t.co/r0bxBa8Lyr
https://twitter.com/i/web/status/1679013217182396419
[ 0.007564250845462084, 0.00746985524892807, -0.051326148211956024, 0.008678121492266655, 0.0006509377853944898, -0.015417981892824173, -0.010232505388557911, 0.027412543073296547, 0.013114724308252335, -0.03627316281199455, 0.007023048121482134, -0.07717801630496979, -0.060463663190603256, ...
1678754237520461829
This for sure will make it into our @gisday activities! https://t.co/RMSn3zjRgY https://t.co/8US8Rc7zp3
https://twitter.com/i/web/status/1678754237520461829
[ -0.024235114455223083, 0.03250043839216232, 0.024574583396315575, 0.010051226243376732, -0.024294152855873108, -0.029282866045832634, 0.03710540756583214, -0.004988714121282101, 0.0034057567827403545, -0.008855706080794334, -0.01889217272400856, 0.014235546812415123, -0.02148984558880329, ...
1678879573486759937
Recording of the 4-hour #geemap workshop at the SciPy Conference 2023 YouTube: https://t.co/oXeTT9dPgZ Notebook: https://t.co/Bl5VFOJOEJ #geospatial #EarthEngine #opensource #dataviz #SciPy2023 https://t.co/YfMuU2fBvZ
https://twitter.com/i/web/status/1678879573486759937
[ 0.003245299682021141, 0.010347149334847927, 0.030020596459507942, -0.01977427303791046, -0.032263949513435364, -0.006849787663668394, 0.05278180539608002, -0.0018857396207749844, -0.009937549009919167, -0.03135652467608452, 0.050286389887332916, -0.0019109458662569523, 0.0073854196816682816,...
1678777783848419330
We're launching Keras Core, a new library that brings the Keras API to JAX and PyTorch in addition to TensorFlow. It enables you to write cross-framework deep learning components and to benefit from the best that each framework has to offer. Read more: https://t.co/xmmxBfSZgh https://t.co/k5K22UZNdR
https://twitter.com/i/web/status/1678777783848419330
[ -0.012916064821183681, -0.021053532138466835, 0.03194965794682503, -0.01630956120789051, 0.07128652185201645, 0.00817786529660225, -0.041853126138448715, 0.0465162992477417, -0.03215742111206055, -0.020268641412258148, -0.017821630463004112, -0.049171075224876404, -0.015316905453801155, 0....
1678691522630549506
Training YOLOv5 object detection models has never been this easy. Train your next model in 3 simple steps: 1. Download an object detection dataset from @roboflow Universe or use their annotation tool to create a custom dataset. Download in the YOLOv5 format. 2. Train the… https://t.co/daEfnROrIz
https://twitter.com/i/web/status/1678691522630549506
[ -0.0063998084515333176, 0.00462921429425478, -0.038433853536844254, -0.04247092828154564, 0.004605878610163927, -0.016848357394337654, 0.03540021926164627, 0.020500389859080315, -0.006884024012833834, 0.02092043124139309, -0.0007839328027330339, -0.0403473824262619, 0.0065981619991362095, ...
1678435054215655426
The Right Way to Run Shell Commands From Python #python https://t.co/RvvcjFeyHL
https://twitter.com/i/web/status/1678435054215655426
[ -0.021897632628679276, 0.005667129997164011, 0.02958414889872074, 0.02029999904334545, 0.033405061811208725, -0.013294991105794907, 0.045270003378391266, 0.03387429937720299, 0.004974449519068003, 0.06260935217142105, 0.004217528738081455, -0.028913812711834908, -0.006898871622979641, -0.0...
1678071435011534850
"That is one of the great shots we've seen here in YEARS!" @AndreyRublev97, take a bow! 😱 #Wimbledon https://t.co/uEHcbcf1k8
https://twitter.com/i/web/status/1678071435011534850
[ -0.014985952526330948, 0.0007311590597964823, -0.03161560744047165, 0.0465887188911438, 0.0030048952903598547, -0.007409506943076849, -0.02057197503745556, 0.018581554293632507, -0.03169265761971474, -0.02902163751423359, 0.0032344358041882515, -0.017541397362947464, 0.007736963219940662, ...
1678317174354575361
screencapture - take screenshots taskpolicy - control scheduling of processes say - text-to-speech engine pmset - configure power management networksetup - configure network settings softwareupdate - manage OS updates system_profiler - view system information
https://twitter.com/i/web/status/1678317174354575361
[ -0.020280681550502777, 0.002046449575573206, 0.04761977866292, 0.005738023668527603, -0.037840936332941055, -0.019508669152855873, -0.008976193144917488, 0.04735018312931061, 0.010777559131383896, -0.012915914878249168, 0.050977423787117004, -0.011126803234219551, -0.009803351014852524, 0....
1677964429894320131
Vive @LeTour! Vive la France! 🇨🇵 #GPSArt #StravaArt https://t.co/a668NspZxz
https://twitter.com/i/web/status/1677964429894320131
[ 0.023746438324451447, 0.010041959583759308, -0.014064070768654346, -0.014450300484895706, -0.014090707525610924, -0.007598061114549637, 0.000564777001272887, 0.0076979477889835835, -0.030445517972111702, -0.02575749345123768, 0.014836529269814491, 0.0025171490851789713, -0.05044952780008316,...
1678294912138113024
Quick hacks on GitHub Pages https://t.co/hXBiVnWaNj
https://twitter.com/i/web/status/1678294912138113024
[ -0.02991078421473503, 0.00870700553059578, -0.006873952224850655, -0.01436619646847248, -0.0015848278999328613, -0.04672832787036896, -0.02780131809413433, 0.006379318423569202, -0.023669671267271042, 0.004037083126604557, 0.02387334406375885, 0.012751363217830658, 0.003673381870612502, 0....
1678080934589997060
Great article by @eugeneyan on the architecture of several big players search and recommendation system: https://t.co/LRmcOWD0N5
https://twitter.com/i/web/status/1678080934589997060
[ -0.0019721982534974813, -0.0206187441945076, 0.003081349190324545, 0.014105431735515594, 0.04744064062833786, 0.012116377241909504, -0.05194466933608055, 0.05151314660906792, -0.06564554572105408, -0.005481699947267771, 0.03581646829843521, -0.06267882138490677, -0.0024273209273815155, -0....
1678071778890117121
What are the best Python-based LLMs for running on an M1/M2 Mac right now? I'm particularly interested in models with weights that are available to download without having to click through a license or similar
https://twitter.com/i/web/status/1678071778890117121
[ 0.0154338413849473, 0.036698244512081146, 0.03712036833167076, -0.001968804281204939, -0.0015606977976858616, -0.01633085124194622, 0.0043366458266973495, 0.05756691098213196, 0.01282196119427681, 0.023216718807816505, -0.005273229442536831, -0.006631935015320778, -0.05524523928761482, 0.0...
1678060204943097863
When Deepmind needs semantic retrieval, they just use the largest index on the planet. Fun fact: Query-doc similarity was done via simple TF-IDF instead of vectors. It performed better than vector retrieval when retrieve docs > 45 (they used 50). https://t.co/7Y7urCaA95 https://t.co/UdSGaEIDqi
https://twitter.com/i/web/status/1678060204943097863
[ 0.0014212580863386393, 0.017913756892085075, 0.02008434757590294, 0.0046891141682863235, 0.033376023173332214, 0.01567932590842247, -0.021003656089305878, 0.057150375097990036, -0.049719296395778656, -0.018386179581284523, 0.003699580207467079, -0.03235457092523575, -0.005828674416989088, ...
1677805381089730563
Happy to announce the 0.11.0 version of imbalanced-learn. This release makes sure to be fully compatible with scikit-learn 1.0+ with a couple of new improvements. pip install -U imbalanced-learn conda install -c conda-forge imbalanced-learn Visit: https://t.co/X5yHQGkFXZ
https://twitter.com/i/web/status/1677805381089730563
[ 0.031701892614364624, -0.03465642035007477, 0.053871650248765945, -0.007348584942519665, -0.03590724617242813, -0.04636671394109726, -0.014600123278796673, 0.029264943674206734, -0.040932103991508484, 0.04688429459929466, 0.01628226414322853, -0.03996163606643677, -0.06405076384544373, -0....
1677972333825277954
Here is a link to a Google Colab with the code: https://t.co/F3Gtd6JUGQ Only thing you need to do is to use your OpenAI key.
https://twitter.com/i/web/status/1677972333825277954
[ 0.022908860817551613, 0.012391611002385616, 0.0354740247130394, -0.02199482172727585, 0.016510577872395515, -0.05026065185666084, 0.005735892802476883, 0.0066354707814753056, -0.007572651375085115, -0.029411274939775467, 0.02422785572707653, -0.03218810632824898, 0.004373509902507067, -0.0...
1677727901842821122
You may have seen this, but I bet all your followers haven’t yet. They should. https://t.co/AxAjK0EAGQ
https://twitter.com/i/web/status/1677727901842821122
[ 0.016309557482600212, 0.01708895154297352, -0.054990630596876144, 0.04456983134150505, -0.008414576761424541, 0.008522826246917248, 0.02470969967544079, 0.05216171592473984, 0.021765319630503654, 0.011568238958716393, 0.006433614995330572, -0.036516088992357254, 0.0004686291213147342, 0.00...
1677559271330856960
@arpagon @jerryjliu0 This is the way to do it
https://twitter.com/i/web/status/1677559271330856960
[ -0.011300248093903065, -0.0457846075296402, 0.008922016248106956, 0.02843664586544037, 0.007152932696044445, -0.05602705106139183, -0.03948156163096428, -0.03110668621957302, 0.020193083211779594, 0.01720205508172512, -0.017114512622356415, -0.012759285978972912, 0.006496365647763014, -0.0...
1677306768769122304
Need a color palette? https://t.co/So75Vmw9s1 https://t.co/X1NiTcb7Wd
https://twitter.com/i/web/status/1677306768769122304
[ -0.02211722731590271, -0.03345484286546707, 0.0008623689063824713, -0.011128364130854607, 0.004229412414133549, -0.08872254192829132, 0.02665734477341175, -0.013239899650216103, 0.0019736201502382755, 0.04025233909487724, -0.014926591888070107, -0.02098853886127472, 0.017462970688939095, 0...
1677277204793921537
"An Introduction to Statistical Learning" in Python 🐍 A free pdf version is also available: https://t.co/YIhWmfwo9b https://t.co/ModE7yxLg3
https://twitter.com/i/web/status/1677277204793921537
[ -0.028995217755436897, -0.02901638112962246, -0.03430747985839844, -0.018825726583600044, -0.007217057514935732, -0.03917529061436653, 0.03324925899505615, 0.018942130729556084, -0.014010827988386154, 0.03265665844082832, 0.0017976505914703012, -0.03509056195616722, -0.027619531378149986, ...
1676980600656494594
Goodhart's law is very real. Reminded again of this super excellent post from @jaschasd on applying technical machine learning techniques to mitigate societal/product overfitting: https://t.co/yRma5gQqW1 https://t.co/mNYVHyoxO6
https://twitter.com/i/web/status/1676980600656494594
[ -0.01961515098810196, -0.0051244390197098255, -0.0029616747051477432, 0.04011290892958641, 0.04133029654622078, -0.01371082291007042, -0.014349951408803463, 0.03378249332308769, -0.04787375032901764, -0.02288687974214554, 0.010446703992784023, -0.05813023820519447, -0.04300420358777046, 0....
1676748983535013888
My favorite feature in @LangChainAI are all the DocumentLoaders because it is a convenient data acquisition tool even outside of LLMs Some notes: https://t.co/7cIgS56PWb
https://twitter.com/i/web/status/1676748983535013888
[ 0.016805054619908333, 0.0007814851123839617, 0.017578622326254845, -0.0016221546102315187, 0.030995991080999374, -0.03227637708187103, -0.012777176685631275, 0.016418272629380226, -0.025701064616441727, 0.01151012908667326, -0.0033176648430526257, -0.02028610184788704, -0.024140596389770508,...
1676607003643224065
The other suggestion was to add types and docstrings. Personally don't think types buy us anything here but more typing 😉 Go crazy with docstrings and comments if you want. I have nothing against docstrings. https://t.co/KrBIPvGPMM
https://twitter.com/i/web/status/1676607003643224065
[ -0.016690030694007874, 0.027096932753920555, 0.0005998213891871274, -0.038088660687208176, 0.015310441143810749, -0.001835078583098948, -0.009499671868979931, -0.019644150510430336, 0.01672002114355564, 0.010579350404441357, -0.013068608939647675, -0.00836001057177782, 0.014193274080753326, ...
1676606993123930115
The next advice is to remove lambdas. The complaint is that they make it hard to debug. I don't agree with a "no lambda" policy. Removing them really complicates the comprehensions but ... 🤷‍♀️ https://t.co/c81xyD4oS3
https://twitter.com/i/web/status/1676606993123930115
[ 0.0472169853746891, 0.01109449751675129, 0.034486327320337296, -0.006716466508805752, 0.04180794954299927, 0.016824785619974136, -0.04748594015836716, -0.042166560888290405, 0.052745554596185684, 0.05107204243540764, 0.021501658484339714, -0.05017551779747009, -0.011669767089188099, 0.0086...
1676582601291595777
A little known fact is that conformal prediction has been originally developed for Support Vector Machines and the first published paper was called 'Learning by Transduction' (as Conformal Prediction was then called from later 1990s until 2005). Support Vector Machines was the… https://t.co/K9eT5K3SDN
https://twitter.com/i/web/status/1676582601291595777
[ 0.016786819323897362, -0.012543724849820137, -0.0013723564334213734, 0.013174622319638729, 0.012976692989468575, 0.010830405168235302, 0.014646715484559536, -0.009451090358197689, -0.04839354008436203, 0.02615131437778473, 0.0008381038787774742, -0.0364188626408577, 0.012580836191773415, 0...
1676674903406149632
Makefile Tricks for Python Projects #python https://t.co/7txaNDzrYF
https://twitter.com/i/web/status/1676674903406149632
[ -0.015394289046525955, -0.012465391308069229, 0.023372607305645943, -0.01933072879910469, -0.00431719608604908, -0.015406005084514618, 0.014269592240452766, 0.016964178532361984, -0.02436843328177929, 0.03184298053383827, -0.003904221346601844, -0.007187516428530216, -0.004797535482794046, ...
1676606977680572418
I generally ask folks how they would write this code and I get advice but no actual takers on writing it "better". So I've gone through the suggestions... Remove "magic literals". I'm ok with this one. Here's before and after: https://t.co/89x3Z97dAj
https://twitter.com/i/web/status/1676606977680572418
[ 0.008583419024944305, 0.009861464612185955, 0.003402205416932702, -0.030751992017030716, 0.008536083623766899, 0.005084571428596973, -0.038625385612249374, -0.037962693721055984, 0.04171794280409813, 0.022736594080924988, 0.047966163605451584, -0.05598156526684761, -0.022389469668269157, 0...
1676253871868395523
Oh and by the way, InstructBLIP directly comes with 4-bit and 8-bit inference thanks to the magic of @Tim_Dettmers, enabling same performance and reducing the amount of memory by half or more!! Hat/tip to @younesbelkada for integrating 💪 https://t.co/XOq6Q09a9V https://t.co/bbxs7HJYB0
https://twitter.com/i/web/status/1676253871868395523
[ -0.019576534628868103, -0.002722180914133787, -0.01128691527992487, 0.026845337823033333, 0.0404851995408535, -0.012365651316940784, 0.016318606212735176, 0.005078749731183052, -0.036981116980314255, -0.008731250651180744, 0.0456978864967823, 0.00782626960426569, 0.0420200489461422, 0.0213...
1676357756058468352
🐍📰 The minimax algorithm can be used to find optimal strategies in many different games. In this tutorial, you'll learn to implement minimax in Python while playing the game of Nim. You'll also learn how you can make the algorithm more efficient #python https://t.co/x1to1viVH8
https://twitter.com/i/web/status/1676357756058468352
[ -0.015936732292175293, 0.003865782404318452, 0.02877136692404747, -0.020968761295080185, 0.03253651037812233, -0.024165580049157143, 0.007844045758247375, 0.03301011025905609, -0.03222866728901863, 0.044258177280426025, 0.013296397402882576, -0.053185589611530304, -0.06682534515857697, 0.0...
1676270163283386371
It's the 100th Data Vis Dispatch! We're celebrating just the way we like to — with lots of great charts and maps on everything from gender equality to hot dogs to the Loch Ness Monster 🥂 The party continues on our blog: 🎉 https://t.co/1T5ekOvlvS https://t.co/6nwrvgtZFm
https://twitter.com/i/web/status/1676270163283386371
[ 0.03905594348907471, 0.02360668033361435, 0.03362640365958214, -0.0024967342615127563, 0.0311345886439085, -0.02464275248348713, -0.030872292816638947, 0.02333126962184906, -0.05933145806193352, -0.02118043787777424, 0.006154130678623915, -0.01084595825523138, -0.02188863977789879, -0.0249...
1675982820706226176
one of the missing pieces of https://t.co/AY9kjOvQ7I falls into place - a python script to ship files in a folder to azure blob. https://t.co/r6hO6W0D0q
https://twitter.com/i/web/status/1675982820706226176
[ -0.03374696895480156, -0.002773699816316366, 0.029411228373646736, -0.008630055002868176, 0.044296346604824066, -0.05086899921298027, 0.023335669189691544, 0.025752084329724312, -0.02883128821849823, 0.016431625932455063, 0.027864722535014153, -0.017094412818551064, -0.018917081877589226, ...
1675936797996511246
Learning so much from Andrew Kelley's "Practical Data-Oriented Design" talk. But also realizing how much I've learned since I started working on Ruff :) https://t.co/39YzQAxrsv
https://twitter.com/i/web/status/1675936797996511246
[ -0.04059150815010071, -0.015236067585647106, 0.014651709236204624, 0.019853921607136726, 0.011836813762784004, 0.010988782159984112, -0.011359350755810738, 0.02712276577949524, 0.013176561333239079, -0.011737045831978321, 0.0620274692773819, -0.023459838703274727, -0.021606996655464172, -0...
1675809284410441728
*RGB no more: Minimally-decoded JPEG Vision Transformers* by @jespark0 @jcjohnss #CVPR2023 You can directly train a Vision Transformer on JPEG data (no decoding), achieving SOTA results with much faster inference on JPEG inputs. https://t.co/t76ug6fL5U https://t.co/PKU1A0Smc6
https://twitter.com/i/web/status/1675809284410441728
[ 0.0038271762896329165, 0.0023897599894553423, -0.06332345306873322, 0.004821055568754673, 0.0022888886742293835, -0.024197248741984367, 0.03814122453331947, 0.0234970822930336, -0.051551174372434616, 0.010490616783499718, 0.04027732089161873, 0.016744637861847878, 0.015356173738837242, 0.0...