PR #2887 fix mlu device longTensor bugs

fix mlu device longTensor bugs #2887

muellerzr merged 34 commits into huggingface:main from huismiling:main

Add Cambricon MLU accelerator support

bc5ccfb4

up mlu support for test

3ad38dcd

fix mlu device MULTI_MLU

be32c91c

Update src/accelerate/utils/imports.py

421c1424

up mlu for quality check

78cd1cb6

fix mlu device longTensor error

3abd0383

fix mlu device tensor dtype check

0542987f

fix mlu device send_to_device with torch dynamo error

e024276c

Refactor AcceleratorState

a50b0d91

Should be near complete now

ff628be7

Last missing piece

a1aac83c

Make my way to the acceleratorstate

31ea8cc5

Include update to global var

47b60ca6

Don't use global

2082a9a2

gpu -> cuda

26c484e2

Don't use update for dict, easier to read

5ac5d566

Fix tests

2baa5c3b

stash

d709f666

Getting closer...

ac243159

Needed to spawn at the very end after env was setup

16288982

Explain set_device before deepspeed

6958e1bb

Make docstring more accurate

2b9d339c

Early return insteaD

194db939

Delineat blocks

31201d3d

Make prepare_backend return state + backend for clarity/less magic

eef1aa0b