webmachinelearning/model-loader

Issues

• Rough ETA/roadmap?

  #36 opened 2 years ago by josephrocca
  8

• Request rename MLDevicePreference to MLDeviceType to align with WebNN API

  #41 opened 2 years ago by BruceDai
  0

• Make MLTensor strongly typed?

  #39 opened 2 years ago by wacky6
  4

• TAG early design review interest?

  #3 opened 4 years ago by anssiko
  2

• Sync and ASync versions of function interfaces

  #38 opened 2 years ago by yuhonglin
  0

• Open question: how to version this API

  #25 opened 2 years ago by yuhonglin
  4

• WebNN's WPT test seems blocking the CL submission of Model Loader API

  #31 opened 2 years ago by yuhonglin
  6

• DevicePreference: "GPU" => "GpuRequired" and "GpuPreferred"

  #35 opened 2 years ago by yuhonglin
  0

• Open Question: Add async/streaming interface?

  #26 opened 3 years ago by yuhonglin
  1

• Add performance/usage Hints

  #34 opened 2 years ago by yuhonglin
  1

• Integration of media capture transform

  #33 opened 2 years ago by yuhonglin
  0

• "default" v.s. "auto" in MLDevicePreference and MLPowerPreference

  #30 opened 2 years ago by yuhonglin
  3

• Open Question: best name of the model loaded by this API

  #29 opened 2 years ago by yuhonglin
  4

• Standard model format

  #28 opened 3 years ago by jbingham
  2

• Web-standard ML model format

  #15 opened 3 years ago by josephrocca
  6

• Open question: should we enable the user to explicitly control whether and what hardware accelerator is used?

  #27 opened 3 years ago by yuhonglin
  0

• Open question: support for non-IEEE 754 float point types

  #23 opened 3 years ago by wacky6
  1

• Is there any code that can be downloaded/imported?

  #5 opened 3 years ago by kirill-konshin
  12

• Consider using Permission Policy to control feature availability

  #22 opened 3 years ago by wacky6
  0

• Use spec-prod to publish spec drafts

  #9 opened 3 years ago by jyasskin
  4