A Method for Layer Bit-Width Allocation in LLM Quantization via Performance Maximization Under a Quality-Degradation Constraint

A new arXiv paper (2608.28003v1) proposes a layer bit-width allocation method for Gemma-3-1B that maximizes latency reduction under a quality-degradation constraint, achieving an 11.0% latency reduction with negligible quality loss (98.90% Top-1 agreement, +0.85% perplexity degradation) on an RTX 5090. The method, built on the prior SA-PTQ sensitivity profile and implemented in TensorRT-LLM, outperforms uniform quantization approaches like GPTQ and AWQ, and can reach up to 19.1% speedup with acceptable quality loss.

arXiv:2608.28003v1 Announce Type: new Abstract: This paper proposes a layer bit allocation method for Gemma-3-1B, formulating the problem as performance maximization latency decrease given a degradation budget constraint allowable level of generation quality loss . This approach is different from time- and resource-consuming uniform layer quantization methods that are used in the literature like GPTQ or AWQ or allocation methods without proven performance-accelerating effect like MixLLM or TorchAO . The layer sensitivity profile resulting from our prior work SA-PTQ is applied using the activation pass-through mode inside TensorRT-LLM. For each layer precision is determined individually in blocks, according to a grouping introduced in the prior step 5+5, 10+10, all26 , differentiating the contribution of FFN, Attention, and lm head to the overall speedup. The clock speed was measured for 13 W8A8 variants on an RTX 5090. We find that for FFN and lm head the time cost of quantization/dequantization is compensated for by the use of integer arithmetic, while for short context lengths, the opposite holds true for Attention: an additional step of quantization slows execution down. We propose a manual implementation of SmoothQuant for TensorRT-LLM which was necessary due to export failures, unavailable for lm head. The best solution found under joint consideration of all three criteria with minimal degradation was FFN 5+5 with lm head, providing an 11.0% reduction in latency with negligible quality loss 98.90% Top-1 agreement, +0.85% perplexity degradation . With acceptable quality loss for FFN all26 + lm head, a speedup up to 19.1% was found possible. We suggest further optimizations: fused attention kernels in INT8, KV-cache quantization, using FP8 instead of INT8 and partial Attention quantization analogous to FFN.