QATMethod

Enums · AiDotNet.Enums

Specifies the Quantization-Aware Training (QAT) method to use during model training. QAT simulates quantization effects during training so the model learns to be robust to low precision.

For Beginners

Normally, we train a model in full precision (32-bit) and then compress it afterward (PTQ). QAT is smarter - it simulates the compression DURING training, so the model learns to work well even when compressed.

How It Works

Analogy: It’s like training for a marathon at high altitude - when you compete at sea level, you perform better because you trained under harder conditions.

QAT vs PTQ Comparison:

Research References:

Fields

Field Summary
QATMethod EfficientQAT EfficientQAT - memory-efficient QAT optimized for large language models.
QATMethod ParetoQ ParetoQ - state-of-the-art QAT achieving optimal accuracy across all bit widths.
QATMethod QABLoRA QA-BLoRA - Quantization-Aware fine-tuning with Balanced Low-Rank Adaptation.
QATMethod Standard Standard QAT using Straight-Through Estimator (STE) for gradient propagation.
QATMethod ZeroQAT ZeroQAT - zeroth-order optimization based QAT that doesn’t require backpropagation.
int value__