Scribd
Upload
72 views3 pages

LLM A7

LLM Assignment 7

Uploaded by

shanthidl
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
72 views3 pages

LLM A7

LLM Assignment 7

Uploaded by

shanthidl
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 3
Assessment submited. x (hitps://swayam.gov.in) @ miatneren) NPTEL (hips:tswayam.govinlexplorer?ncCode=NPTEL) » Introduction to Large Language Models (LLMs) (course) Course outline ‘About NPTEL 0 How does an NPTEL online course work? 0 Week 10 Wook 20) Week 3 0 Thank you for taking the Week 7 7. Week 7 : Assignment 7 Ot encoder alone suicens Ot shares vocabulary wih summarization datasets Ottuses a targer context window than BE @ ttakoacy contains a generate Secoder tained jointly ding pe-rainng, 2) For preraning of encaderdecader models which statement) isae ve? “Te encoder ates bisectonaly tits whole inp. ‘gopataiahshanthi@bmsitin ~ : Assignment 1 point 2points Asvessnbarhitfhited. x Week 50 Week 60 Wook 70) ee 18: Prectraning Staiopes: ELM, BERT (unt? ebelesons64) Lee 19: Precrning Statoges: Encoder. ecedor an Docode ony Magee une? lis B2Bleson=68) 2620: atone Huppnaace unt? ntebaelsson=t8) Lecture Mati ant? tins 6asleson-€7) Feodoack Fam uni? ne=Baslseon=68) ‘@-oux: Week? ‘Assignment Year 2025 July Solution “Tne decoder condtens on ear decoder okena and encoder ouput, Unabotd tot is tured into a supervises task via a nosing schome, 2poins (causal mask Cruise mask CPretictm mase Catone asove None ofthe above 4) TS experiments showod that chan and compact pre-taring data can ouperfoem a larger but nsisiercopus primarily point © Larger corpora overt © Noise foros he model to waste capacty on modeling iralevant pats clean data nas longer documents compact data aos bigger batches. 15) What makas samping om an autoregressive language moda straightforward? 1 point OTe modelis dota. OTe voeabulry sal (@ ach conctonal distribution over he vocabulary sradly normalised and can be sampled toker-byokan Beam searen guarantees opty. 15}, Why does ELMo bili input oken roprsertatons rom a charatorJovel GN instad of xo word omibeckings? 1 point O oreducetrinng tne by sharing parameters (@ Te avo UNK tokens and generate presentations fo any sing Oo compress embedsings to 128 mensions (O Te ensure the sare vector fora wor in every context Assessment submited. x n The eum function in num is Used as @ generalized operation for performing tensor 28 39 muitipicatons. Now, considertwo matrices: = [2 S]ané # = [2 9] . Then, what Is the output of the following numpy operation? pumpy.einsum('ij,ij->', a, B) a 2ponts ‘You may submit any numberof ines before the due dat, The nl submission wil be considere for grading ‘Submit Anows

You might also like