Ikibazo gihuza ibibazo · Hagati Urwego
Ibikurikira-Byabajijwe Ikibazo
Ongera usubiremo ururimi-rwerekana amahugurwa, imiterere yabitera, guhatira abarimu no gutandukanya amahugurwa nibisekuru.
Ikibazo 1 Bya 8
Niki gikurikira-ikimenyetso cyerekana ubuhanuzi gisohoka kuri buri ntambwe?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Niki gikurikira-ikimenyetso cyerekana ubuhanuzi gisohoka kuri buri ntambwe?
- Ikwirakwizwa rishoboka hejuru yamagambo kubikurikira
- Ijambo rimwe rihamye
- Ubusobanuro bw'interuro
- Ikibonezamvugo
- Ni ubuhe bwoko bwo kwitondera GPT-yuburyo bwa decoder ikoresha?
- Ibyerekezo byombi
- Impamvu (mask) kwitondera ihisha ibimenyetso bizaza
- Nta kwitabwaho na gato
- Kwitonda bisanzwe
- Igisekuru 'autoregressive' gisobanura iki hano?
- Icyitegererezo gitanga ibimenyetso byose icyarimwe
- Icyitegererezo gisubira inyuma
- Buri kimenyetso gishya cyongeweho kandi kigaburirwa inyuma kugirango uhanure ibizakurikiraho
- Tokens ikorwa muburyo butunguranye
- Ni ubuhe buryo bwo gutakaza busanzwe bugabanywa mugihe cy'amahugurwa?
- Ikigereranyo cya kare
- Gutakaza Hinge
- Intera
- Kwambukiranya-entropiya hejuru yikimenyetso gikurikira
- Kuzamura ubushyuhe mugihe cyo gutoranya bikora iki?
- Bituma umusaruro usohoka kandi udasanzwe
- Ikora ibisohoka rwose
- Yongera ingano yamagambo
- Kuraho igikoresho cya softmax
- Mbere yo guhanura, ni gute inyandiko mbisi isanzwe itegurwa?
- Byahinduwe mu Cyongereza
- Igabanijwemo ibimenyetso na tokenizer
- Yahinduwe amajwi
- Hasigaye nkinyuguti mbisi nta gutunganya
- During next-token training, what is teacher forcing?
- The model receives future target tokens before the input
- The model randomly changes the tokenizer at every step
- The model receives the ground-truth preceding tokens when predicting the next token
- The model always samples its own previous prediction during training
- What mismatch can arise because generation feeds back model-produced tokens?
- The model can no longer use attention
- The vocabulary disappears after the first token
- The model’s loss is computed only on images
- Errors in generated prefixes can affect later predictions, unlike training on the ground-truth prefix