huggingface/accelerate

Pull Requests Commits

Fix CPU error always being raised (#1175)

muellerzr committed 2 years ago

Verified 3ca8c9a9

Fix accelerate test with new config_file errors (#1169)

sgugger committed 2 years ago

Verified 2f83b1af

Fix backward compatibility in configs wrt dynamo backend (#1168)

sgugger committed 2 years ago

Verified b0591c66

sgugger committed 2 years ago

Verified d9871c0f

Remove outdated command directions and use in tests (#1166)

muellerzr committed 2 years ago

Verified abc2beb4

Fix what files get deleted through `total_limit` (#1165)

muellerzr committed 2 years ago

Verified 8749b4ec

Document skip_first_batches in the checkpoint usage guides (#1164)

muellerzr committed 2 years ago

Verified 4a3eaee6

[`Accelerator`] Fix issue with 8bit models (#1155)

younesbelkada committed 2 years ago

Verified 3533e2b0

Attempt to fix import error when PyTorch is build without `torch.distributed` module (#1108)

mfuntowicz committed 2 years ago

Verified 3e0ceac7

Let GradientState know active dataloaders and reset the remainder (#1162)

muellerzr committed 3 years ago

Verified 03b617b6

update support for torch dynamo compile (#1150)

pacman100 committed 3 years ago

Verified 840bb1ae

Fp8 integration (#1086)

sgugger committed 3 years ago

Verified 1bfde6b9

:memo: add a couple more trackers to the docs (#1158)

nateraw committed 3 years ago

Verified 3482495b

Load custom state to cpu (#1156)

Guangxuan-Xiao committed 3 years ago

Verified 947b2a88

Solve arrow keys being environment dependant for accelerate config

p1atdev committed 3 years ago

Verified cac1ed41

[Safetensors] Relax missing metadata constraint (#1151)

patrickvonplaten committed 3 years ago

Verified 9dc5b349

Include a note in the gradient synchronization docs on "what can go wrong" and show the timings (#1153)

muellerzr committed 3 years ago

Verified 0aae1e93

Fixed typos in notebook (#1146)

SamuelLarkin committed 3 years ago

Verified 78151f87

FSDP enhancements and fixes (#1145)

pacman100 committed 3 years ago

Verified 853823d0

fix partial state (#1144)

pacman100 committed 3 years ago

Verified 77ae51a0

Fix notebook_launcher (#1141)

muellerzr committed 3 years ago

Verified ad9cf788

fsdp bf16 enable autocast (#1125)

pacman100 committed 3 years ago

Verified 5f9cea4c

fix lr scheduler issue (#1140)

pacman100 committed 3 years ago

Verified 96ffd349

fix ds dist init kwargs issue (#1138)

pacman100 committed 3 years ago

Verified d88bbbd0

deepspeed dataloader prepare fix (#1126)

pacman100 committed 3 years ago

Verified 075b5d61

Fix multinode with GPU ids when each node has 1 (#1127)

muellerzr committed 3 years ago

Verified 9b5877d1

Expand warning and grab all GPUs available by default (#1134)

muellerzr committed 3 years ago

Verified 586941d1

Add tee and role to launch (#1132)

muellerzr committed 3 years ago

Verified e1b84bf5

[Big model loading] Correct GPU only loading (#1121)

patrickvonplaten committed 3 years ago

Verified b2ea1c7b

Refactor `launch` for greater extensibility (#1123)

Yard1 committed 3 years ago

Verified bdd93cd9

Newer Older