HugeCTR A100 PREVIEW

HugeCTR is a high-efficiency GPU framework designed for Click-Through-Rate (CTR) estimation training, and the new released NVIDIA A100 GPU has excellent acceleration on various scales for AI, data analysis and high performance computing (HPC), and meet extremely severe computing challenges. To demonstrate HugeCTR’s performance on A100 GPU, this version is developed to leverage new features of the latest GPU.

HugeCTR A100 PREVIEW

v2.1_a100

a100-preview

HugeCTR V2.1

New Features:

• Supporting three important networks: Wide and Deep Learning (WDL), Deep Cross Network (DCN) and DeepFM.
• A new embedding implementation LocalizedSlotSparseEmbedding which reduces the memory transactions across GPUs and nodes resiliently to the number of GPUs.
• Supporting multiple Embeddings in one network.
• Supporting dense feature input.
• Supporting new layers like: Dropout / Split / Reshape / Multiply / FmOrder2 / MultCross / Add
• Check bits in data reader to enable data check and error skip.