AutoRound can be combined with several algorithms before (or during) quantization. This page summarizes each combination and rates it along two dimensions:
- Accuracy Gain — does the transform improve the accuracy of the quantized model compared with plain AutoRound?
- Deployment — can the resulting model actually be deployed/served today (kernel support, export path, real inference engine)?
| Light | Meaning |
|---|---|
| 🟢 | Good — clear benefit / ready to deploy |
| 🟡 | Partial — conditional benefit / limited or experimental support |
| 🔴 | Poor — no measurable benefit / not deployable yet |
Note: A Partial/Poor Accuracy Gain rating may stem from two factors: (1) limitations in our current implementation, and (2) our own internal, subjective evaluation. Both are subject to change as the implementation matures and more benchmarks become available.
| Combination | Accuracy Gain | Deployment | Details | CLI Usage | Comments | Reference |
|---|---|---|---|---|---|---|
| AutoRound + AWQ (activation-aware scaling) | 🟢 | 🟢 | awq_details | --algorithm awq,signround |
Recommended when activations are quantized (e.g., W4A4). | arXiv:2306.00978 |
| AutoRound + Hadamard rotation | 🟢 | 🔴 | rotation_details | --algorithm hadamard,signround |
Especially helpful for INT4 (W4A4) and some MXFP4 scenarios. no production kernel. | arXiv:2404.00456 |
| AutoRound + SpinQuant | 🟡 | 🔴 | rotation_details | Python API only | Learns rotation matrices; higher accuracy at extra training cost. no production kernel. | arXiv:2405.16406 |
| AutoRound + LFQ (logit-aware final-block quantization) | 🔴 | 🟢 | lfq_acc | --enable_lfq |
Refines the final block to lift low-bit generation quality | arXiv:2605.29756 |
| AutoRound + MX Attention (mxfp4 variant) | 🟡 | 🔴 | mxnv_acc | --data_type mx_fp4_rceil_v2 |
Adopt 7.25 as the denominator for scale calculation | arXiv:2607.24377 |
| AutoRound + SVDQuant (low-rank outlier absorption) | 🟡 | 🟡 | svdquant_details | --algorithm svdquant,signround |
Recommended for diffusion models; currently only FLUX is supported. | arXiv:2411.05007 |