1-3 中文大规模多模态预训练模型M6.pdfVIP

  • 2
  • 0
  • 约7.67千字
  • 约 26页
  • 2024-01-31 发布于广东
  • 举报

2021

|DataFunSummit

M6:AChinese

MultimodalPretrainer

杨安达摩院智能计算实验室

Content

•Review

•Approach

•DownstreamApplications

•ScalingtoTrillion-parameterModel

2

Review

Previousworkinlarge-scalemultimodal

pretraining

MultimodalPretraining

•Modelarchitecture:

•Transformer-based

•Single-streamortwo-stre

文档评论(0)

1亿VIP精品文档

相关文档