Awesome GitHub RepositoriesModel Compression

Compresses vision transformers using weight multiplexing.

4-bit floating-point quantization for large language models.

Integer-only quantization for efficient inference.

Data-independent pruning for hierarchical vision transformers.

Optimizes quantization in the loss landscape of transformers.

Compresses models by leveraging low-frequency components.

Accelerates vision transformers via soft token pruning.

Multimodal alignment-guided token pruning for acceleration.

Variation-aware quantization for vision transformers.

Scale reparameterization for post-training quantization.

Efficient compression and customization for attention-based models.

Unified pruning for vision-language transformer models.

Oscillation-free quantization for low-bit transformer models.

Pushes binary vision transformers toward convolutional performance.

Explainable pruning framework for vision transformers.

Combines token pruning and squeezing for aggressive compression.

Accurate low-bit quantization for vision transformers.

Data-free quantization framework for vision transformers.

Unified framework for compressing vision transformer models.

Enables fully quantized vision transformers without retraining.