ДаліНаступний посібник
Інструкція Налаштування
ШІ мови
Мова AI GUIDE
LoRA represents a weight update through small trainable matrices while keeping the original selected weight matrix frozen.
Послухайте цей посібникШІ-розказаний · ~3 min
This can reduce trainable-parameter and task-checkpoint requirements, with the actual savings and task performance depending on the configuration.
Full fine-tuning updates a model's trainable base parameters, with optimizer state and checkpoints contributing to resource requirements. LoRA offers a different parameterization for selected weight updates. The original selected matrix stays frozen, while two trainable matrices form an additive low-rank update. Their inner dimension r controls the allowed update rank and parameter count; it is not a declaration that the whole pretrained matrix has rank r. For a matrix with d rows and k columns, factors with shapes d by r and r by k contain r times d plus k parameters. In a stated 4-by-4 example with rank one, the factors contain four plus four, or eight parameters, compared with sixteen entries in a full unconstrained update. This is an illustrative matrix calculation rather than a universal model-wide percentage. Reducing trainable parameters can reduce optimizer-state and task-checkpoint requirements, but the base computation and relevant activations still exist. A compatible update can also be folded into a dense base matrix before inference. The layer then uses the merged matrix instead of evaluating a separate adapter branch, with dtype and implementation compatibility still relevant. Different task adapters can share a base checkpoint, while serving behavior depends on the system used. QLoRA adds another set of choices: a frozen base stored at low precision, with gradients reaching trainable LoRA parameters through higher-precision computation. Its paper uses 4-bit NormalFloat storage and generally BFloat16 computation. Double quantization compresses the quantization constants themselves, addressing metadata overhead rather than meaning that every adapter gradient is quantized twice.
Comparing an original LoRA adapter with full weight fine-tuning on a defined task.
Storing separate compatible task updates while sharing a pretrained base checkpoint.
Studying QLoRA's frozen low-bit base and higher-precision adapter computation.
Calculating the matrix parameter count for a stated adapter rank and layer shape.
Adaptation comparisons can examine selected layers, rank, initialization, data, precision, and serving configuration. A smaller task checkpoint can be useful without removing the memory needed for the base model or guaranteeing a single-device training setup. Reports should separate trainable-state savings from inference latency and evaluate task quality after any merge or precision change. Shared bases can support multiple compatible task updates, while simultaneous serving introduces its own implementation choices. The method's usefulness is best assessed against a defined baseline and resource budget rather than inferred from the adapter file size alone.
Безплатний розсилок
Три перевірені історії про ШІ щоробочого ранку, написані простою мовою. Безплатно назавжди, без реклами.
Один лист щотижня з понеділка по п'ятницю. Відписатися в один клік. Ми ніколи не продаємо та не передаємо вашу адресу.
Перевір себе
Миттєва зворотна связь на кожну відповідь та сертифікат із перевіреним ідентифікатором, яким можна поділитися, після проходження курсу.
Підтримайте безплатну освіту у сфері ШІ. AI Understanding — це некомерційна організація 501(c)(3) — без реклами, без платних стін, назавжди. Зробіть пожертву
LoRA represents a weight update through small trainable matrices while keeping the original selected weight matrix frozen. This can reduce trainable-parameter and task-checkpoint requirements, with the actual savings and task performance depending on the configuration.
The selected base matrix remains fixed while the factorized update is trained.
Trainable-state and update-checkpoint savings differ from the remaining base computation and activations.
Ранг r — це малий внутрішній розмір, спільний для матриць A і B; більший r дає адаптеру більше можливостей, але більше параметрів для навчання.
The paper separates low-bit base storage from higher-precision computation and adapter learning.
The update can be folded into the base matrix under a compatible implementation; this is not a universal latency claim for every serving setup.
Продовжуйте вчитися
Інші посібники, вибрані для цієї теми
ДаліНаступний посібник
Інструкція Налаштування
ШІ мови