Ikibazo gihuza ibibazo · Hagati Urwego
Warmup na Cosine Annealing Gahunda Ikibazo
Gerageza gususurutsa uruhare rwamahugurwa hakiri kare, inzibacyuho yo kwangirika kwa cosine, nuburyo bwo gusoma ingengabihe y'umurongo ugereranije.
Ikibazo 1 Bya 8
Niyihe ntego nyamukuru yicyiciro cyo gushyushya mugitangira imyitozo?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Niyihe ntego nyamukuru yicyiciro cyo gushyushya mugitangira imyitozo?
- Kugirango uhagarike imyitozo hakiri kare mugihe uburemere butemewe kandi gradients ni urusaku
- Kongera ingano yicyiciro buhoro buhoro
- Guhagarika icyitegererezo cyo hasi
- Gucomeka icyitegererezo kugirango byihuse
- Cosine annealing igabanya igipimo cyo kwiga ukurikije ubuhe buryo?
- Umurongo ugororotse
- Isanduku ya cosine itangira hejuru, yoroshye, kandi irambuye hafi ya zeru
- Intambwe yimikorere igabanuka mugihe cyagenwe
- Icyerekezo cyerekana hejuru
- Ninde optimizer yunguka cyane kubushuhe kuko igereranya ryayo ritizerana hakiri kare?
- Ikibaya cya SGD nta mbaraga
- Gushakisha
- Adamu n'ibitandukanye
- K-bisobanura
- Muri cosine formula, T igereranya iki?
- Ubushyuhe bwikitegererezo bwo gutoranya
- Umubare wa GPU wakoreshejwe
- Ingano yamagambo
- Umubare wamahugurwa yintambwe muri gahunda
- In the guide’s warmup-plus-cosine schedule, what begins after warmup reaches its target rate?
- Cosine decay begins from the peak learning rate
- All model weights are reset
- The schedule switches to a constant zero rate
- Momentum is removed from every optimizer
- A linear warmup rises from zero to its target rate over 200 steps. About what rate is used halfway through?
- The full target rate
- About half of the target rate
- The learning rate is exactly zero
- Twice the target rate
- At the cosine horizon t = T, what multiplier does the guide’s cosine formula produce?
- Imwe
- One-half
- Zeru
- It becomes negative one
- Why use a gradual cosine decay after warmup rather than abruptly setting the rate to zero?
- It guarantees a global minimum
- It increases the model’s parameter count
- It removes the need to check training loss
- It lets the rate ease down so optimization can settle gradually