Code

Hsdp + MoE CI tests#3378

Code

Merged

KuuCi merged 72 commits intodevmosaicml/composer:devfrom hsdp-ci-testsmosaicml/composer:hsdp-ci-testsCopy head branch name to clipboardJun 24, 2024

+59-27

Commits

Commits on Jun 4, 2024

• fold ema fsdp state

  

  v-chen_data

  committed
  eb90fda

• debug

  

  v-chen_data

  committed
  7fb0af8

• debug

  

  v-chen_data

  committed
  a4edf86

• more debug

  

  v-chen_data

  committed
  e0759e6

• keep debugging

  

  v-chen_data

  committed
  2ae40ba

• debug

  

  v-chen_data

  committed
  7a908cc

• sanity check

  

  v-chen_data

  committed
  11207bf

Commits on Jun 5, 2024

• debug

  

  v-chen_data

  committed
  874aedc

• debug

  

  v-chen_data

  committed
  96cf25b

• use ema

  

  v-chen_data

  committed
  a2cb3c0

• debug

  

  v-chen_data

  committed
  25c1008

• debug

  

  v-chen_data

  committed
  09f380f

• debug

  

  v-chen_data

  committed
  9e2ea2c

• debug

  

  v-chen_data

  committed
  e93d5c2

• debug

  

  v-chen_data

  committed
  617b7ae

• debug

  

  v-chen_data

  committed
  3d33a30

Commits on Jun 6, 2024

• more fix

  

  v-chen_data

  committed
  fcecdba

• filename test

  

  v-chen_data

  committed
  65acd3e

• revert test

  

  v-chen_data

  committed
  686a6f3

• fully parameterize

  

  v-chen_data

  committed
  317995a

• hsdp test

  

  v-chen_data

  committed
  a71ff08

• revert testing

  

  v-chen_data

  committed
  7200717

• typo

  

  v-chen_data

  committed
  223169b

Commits on Jun 9, 2024

• Merge branch 'dev' of git.luolix.top-regular:mosaicml/composer into hsdp-ci-tests

  

  v-chen_data

  committed
  a15df57

• typo

  

  v-chen_data

  committed
  15d2671

• hsdp

  

  v-chen_data

  committed
  9650dd9

• split off test

  

  v-chen_data

  committed
  7b1bb0b

• precommit

  

  v-chen_data

  committed
  dff8525

• float to int

  

  v-chen_data

  committed
  c5e716c

• pyright

  

  v-chen_data

  committed
  592c35f

• oom

  

  v-chen_data

  committed
  9111519

Commits on Jun 10, 2024

• print

  

  v-chen_data

  committed
  4676b86

• rm tp

  

  v-chen_data

  committed
  737cb3d

• tp cfg

  

  v-chen_data

  committed
  17645dc

• tp?

  

  v-chen_data

  committed
  a1bff4a

• rm tp line

  

  v-chen_data

  committed
  1f0992f

• type annotation

  

  v-chen_data

  committed
  68e3a9c

• revert

  

  v-chen_data

  committed
  bf85dd8

• readd tp

  

  v-chen_data

  committed
  2ea6479

• type

  

  v-chen_data

  committed
  f84832e

• world size

  

  v-chen_data

  committed
  ddd4fdd

• revert

  

  v-chen_data

  committed
  78f00f1

Commits on Jun 11, 2024

• revert monolithic cpkt + include sharded cpkt

  

  v-chen_data

  committed
  e65e4fd

• enumerate

  

  v-chen_data

  committed
  1571f7e

• precommit

  

  v-chen_data

  committed
  9150fb7

• precommit

  

  v-chen_data

  committed
  a255a6f

• sharded

  

  v-chen_data

  committed
  2796140

Commits on Jun 12, 2024

• sync

  

  v-chen_data

  committed
  9e197f5

Commits on Jun 13, 2024

• only sync on first trainer

  

  v-chen_data

  committed
  3b02940

• typo

  

  v-chen_data

  committed
  e457efb

Commits on Jun 20, 2024

• hsdp

  

  v-chen_data

  committed
  eda5ede

• Merge branch 'dev' into hsdp-ci-tests

  dakinggg

  authored
  4b55781

• Merge branch 'hsdp-ci-tests' of git.luolix.top-regular:mosaicml/composer into hsdp-ci-tests

  

  v-chen_data

  committed
  f8f1145

• xfail

  

  v-chen_data

  committed
  8ed5c33

• explicit sync

  

  v-chen_data

  committed
  60ce09c

• test

  

  v-chen_data

  committed
  f9c7892

• revert test

  

  v-chen_data

  committed
  d605299

• sync, docker issue

  

  v-chen_data

  committed
  97b3005

• pre-commit

  

  v-chen_data

  committed
  4d642ea

Commits on Jun 21, 2024

• sync

  

  v-chen_data

  committed
  fa33327

• pytest

  

  v-chen_data

  committed
  8a192e1

• xfail

  

  v-chen_data

  committed
  bb6150d

• rm world_size param

  

  v-chen_data

  committed
  635f92e

Commits on Jun 22, 2024

• Merge branch 'dev' into hsdp-ci-tests

  KuuCi

  authored
  3c23856

Commits on Jun 24, 2024

• Merge branch 'dev' into hsdp-ci-tests

  KuuCi

  authored
  13ab59c

• im so sorry pls forgive me king

  

  v-chen_data

  committed
  6352972

• Merge branch 'hsdp-ci-tests' of git.luolix.top-regular:mosaicml/composer into hsdp-ci-tests

  

  v-chen_data

  committed
  e5778a3

• Merge branch 'dev' into hsdp-ci-tests

  KuuCi

  authored
  5939696

• the kings comments

  

  v-chen_data

  committed
  2200f73

• Merge branch 'hsdp-ci-tests' of git.luolix.top-regular:mosaicml/composer into hsdp-ci-tests

  

  v-chen_data

  committed
  95be0e3

• Update tests/trainer/test_fsdp_checkpoint.py

  Show description for dd64f47
  

  KuuCi

  and

  mvpatel2000

  authored
  dd64f47

• precommit

  

  v-chen_data

  committed
  124e6e4