Towards Efficient Pre-training: Exploring FP4 Precision in Large Language Models
This work addresses the problem of computational efficiency for researchers and practitioners training large language models, representing an incremental advance in low-precision training methods.
The paper tackles the high computational cost of training large language models by developing an FP4 training scheme that uses mixed-precision quantization strategies for different model components and training stages, achieving accuracy comparable to BF16 and FP8 with reduced theoretical computational cost.
The burgeoning computational demands for training large language models (LLMs) necessitate efficient methods, including quantized training, which leverages low-bit arithmetic operations to reduce costs. While FP8 precision has shown potential, leveraging FP4 remains challenging due to inherent quantization errors and limited representation capability. Based on the Transformer architecture, we present an FP4 training scheme for LLMs, overcoming these obstacles through mixed-precision quantization strategies tailed for different modules and training stages. This allows us to apply the precision level suitable to distinct components within the model, ensuring that multi-head attention and linear layers are handled appropriately. Our pretraining recipe ensures stability in backpropagation by incorporating fine-grained quantization methods with a target precision training schedule. Experimental results demonstrate that our FP4 training scheme achieves accuracy comparable to BF16 and FP8, with smaller theoretical computational cost. With the advent of next-generation hardware supporting FP4, our method sets the foundation for efficient ultra-low precision training.