Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 20, 2022

• Disk cache support in CPU-only mode

  

  Gnome Ann

  committed
  90fd8b1

  Copy full SHA for 90fd8b1

• Add disk cache slider

  

  Gnome Ann

  committed
  47a58a3

  Copy full SHA for 47a58a3

Commits on May 13, 2022

• OPT breakmodel bug fix

  

  Gnome Ann

  committed
  a051bf4

  Copy full SHA for a051bf4

• OPT breakmodel

  

  Gnome Ann

  committed
  defbb53

  Copy full SHA for defbb53

Commits on Feb 1, 2022

• XGLM breakmodel

  

  Gnome Ann

  committed
  e7f65ce

  Copy full SHA for e7f65ce

Commits on Nov 26, 2021

• Don't pin CPU layers after running out of pinned memory

  

  Gnome Ann

  committed
  95aff61

  Copy full SHA for 95aff61

Commits on Nov 25, 2021

• Breakmodel support for GPTJModel

  

  Gnome Ann

  committed
  25c9be5

  Copy full SHA for 25c9be5

• In breakmodel mode, move layers to GPU as soon as model loads

  Show description for f8bcc34
  

  Gnome Ann

  committed
  f8bcc34

  Copy full SHA for f8bcc34

Commits on Oct 6, 2021

• Breakmodel's CUDA stream should be on primary device

  

  Gnome Ann

  committed
  3649ba9

  Copy full SHA for 3649ba9

Commits on Oct 5, 2021

• Slightly increased performance in breakmodel mode

  Show description for f9e6a6d
  

  Gnome Ann

  committed
  f9e6a6d

  Copy full SHA for f9e6a6d

• Multiple GPU support

  

  Gnome Ann

  committed
  a283d34

  Copy full SHA for a283d34

Commits on Oct 2, 2021

• Clarify licensing for breakmodel.py

  

  Gnome Ann

  committed
  0937bb3

  Copy full SHA for 0937bb3

Commits on Sep 24, 2021

• K80 test

  

  Gnome Ann

  committed
  4d9eab3

  Copy full SHA for 4d9eab3

Commits on Aug 28, 2021

• Fix for when breakmodel layers is 0

  

  Gnome Ann

  committed
  b5c28f4

  Copy full SHA for b5c28f4

Commits on Aug 20, 2021

• Fix for non-rotary models without "rotary" in config.json

  

  Gnome Ann

  committed
  8bfcf86

  Copy full SHA for 8bfcf86

• Specifically import torch.cuda.comm in breakmodel.py

  

  Gnome Ann

  committed
  eef0db8

  Copy full SHA for eef0db8

• Implement arrmansa's low VRAM patch

  

  Gnome Ann

  committed
  b1c13f8

  Copy full SHA for b1c13f8

Loading