4 ms·
Fast GPT-2 inference written in Fortran
- bmacho 3y ago2 months ago https://news.ycombinator.com/item?id=35159961 https://news.ycombinator.com/item?id=35159961 FastGPT: Faster than PyTorch in 300 lines of Fortran - 12 comments
- eternalban 3y ago"Ondřej [Certik] is the original author of SymPy, SymEngine and LFortran". https://www.sympy.org/en/index.html https://www.sympy.org/en/index.html https://symengine.org/ https://symengine.org/ https://lfortran.org/ https://lfortran.org/ << "Modern interactive LLVM-based Fortran compiler" ~ Raise your hand if you first encountered Fortran in (eng) school and wrote your first Fortran programs on punch cards. (MTS for bonus).
- StephenSmith 3y agoMy school still made me take it in 2011... for a Mech-E.
- planetis 3y agoYes, besides the punch cards in 2013, superseded by a course in python at around 2016.
- Loic 3y agoI can raise only one half of a hand as I started without punch cards, but we had cards at home because my mother "recycled" the unused one she add at the office to take notes.
- pklausler 3y ago(warily raises hand stained with ink from the diagonal lines drawn across the edges to ease collation of dropped decks)
- eternalban 3y agoWonderful, I think I remember that! A great little hack right there.
- pklausler 3y agoPRINT 10 10 FORMAT(14HYOU'RE WELCOME) END
- mvcalder 3y agoNo punch cards, but I did write the Windows version of ITSM (software accompanying the Brockwell and Davis Time Series book). The while thing was written in Fortran, the event loop and calls to the Windows API for UI and graphics. It's still the best damn visualization of a series + periodogram I've ever come across (if I do say so myself).
- dvh 3y agoReal fortran programmers can write fortran code in any language.
- certik 3y agoThe author here. If you have any questions, let me know. If there is anybody here who wants to help parallelize this, let me know!
- collaborative 3y agoThe benchmark says the fastest model takes 0.3s for 20 tokens. Does this mean it would take 30 seconds for 2000 tokens?
- jdkee 3y agoSo does the model simply extract the most likely answer to the prompt based on model weights?
- lakerkolya2 3y ago[dead]