Exploiting activation sparsity is a promising approach to significantly
accelerating the inference process of large language models (LLMs) without
compromising performance. However, activation sparsity is determined by
activation functions, and commonly used ones like SwiGLU and GeGLU exhibit
limited sparsity. Simply replacing these functions with ReLU fails to achieve
sufficient sparsity. Moreover, inadequate training data can further increase
the risk of performance degradation. To address these challenges, we propose a
novel dReLU function, which is designed to improve LLM activation sparsity,
along with a high-quality training data mixture ratio to facilitate effective
sparsification. Additionally, we leverage sparse activation patterns within the
Feed-Forward Network (FFN) experts of Mixture-of-Experts (MoE) models to
further boost efficiency. By applying our neuron sparsification method to the
Mistral and Mixtral models, only 2.5 billion and 4.3 billion parameters are
activated per inference iteration, respectively, while achieving even more
powerful model performance. Evaluation results demonstrate that this sparsity
achieves a 2-5x decoding speedup. Remarkably, on mobile phones, our
TurboSparse-Mixtral-47B achieves an inference speed of 11 tokens per second.
Our models are available at https://huggingface.co/PowerInfer

利用激活稀疏性是大型语言模型（LLM）推理过程中显著加速的一种有希望的方法，但是激活稀疏性由激活函数决定，而常用的 SwiGLU 和 GeGLU 等函数显示出有限的稀疏性，仅用 ReLU 替换这些函数无法实现足够的稀疏性，此外，不充分的训练数据还会进一步增加性能下降的风险。为了解决这些挑战，我们提出了一种新颖的 dReLU 函数，旨在改善 LLM 的激活稀疏性，以及高质量的训练数据混合比例以促进有效稀疏化。此外，我们利用 Mixture-of-Experts（MoE）模型中 Feed-Forward Network（FFN）专家内的稀疏激活模式进一步提高效率。通过将我们的神经元稀疏化方法应用于 Mistral 和 Mixtral 模型，每次推理迭代只激活 25 亿和 43 亿个参数，同时实现更强大的模型性能。评估结果表明，这种稀疏性实现了 2-5 倍的解码加速。值得注意的是，在手机上，我们的 TurboSparse-Mixtral-47B 每秒实现 11 个标记的推理速度。我们的模型可在 https://huggingface.co/PowerInfer 获得。