microsoft/DeepSpeed

Pull Requests Commits

tjruwase committed 3 years ago

cc13b5aa

zero inference timers

tjruwase committed 3 years ago

583987a9

Add more annotations

Quentin-Anthony committed 3 years ago

8e23f504

Add labels to common collective calls for logger

Quentin-Anthony committed 3 years ago

31c7dcf7

Add more robust arg checks

Quentin-Anthony committed 3 years ago

e55c8e93

Clean up logger

Quentin-Anthony committed 3 years ago

0023b3e1

Add commslogger class, and introduce rough prototype comms logging

Quentin-Anthony committed 3 years ago

2e7129c6

Remove misleading world_size prints

Quentin-Anthony committed 3 years ago

656b4152

Simplify ZeRO stage 1 check for previous commit

Quentin-Anthony committed 3 years ago

09063a3d

Relax assertion to allow Megatron-DeepSpeed MoE to use ZeRO 1

Quentin-Anthony committed 3 years ago

9976681a

Remove unnecessary print and update backend description

Quentin-Anthony committed 3 years ago

f070a0c8

Merge branch 'staging-comms-next' of https://github.com/microsoft/DeepSpeed into staging-comms-next

Quentin-Anthony committed 3 years ago

c06c72d3

Update comments and move custom comm ops to internal functions

Quentin-Anthony committed 3 years ago

54796bb8

Quentin-Anthony committed 3 years ago

Verified 68eb9f4e

revert deepspeed/__init__.py logging calls

Quentin-Anthony committed 3 years ago

977ee324

Delete distributed.py

awan-10 committed 3 years ago

Verified 7f8ca013

Delete stage1.py

awan-10 committed 3 years ago

Verified c93fcfef

Staging comms v1 (#301)

Quentin-Anthony committed 3 years ago

Verified 867a8537

Fairseq support (#1915)

jeffra committed 3 years ago

Verified 50893458

[pipe] prevent deadlock with multiple evals sequence (#1944)

stas00 committed 3 years ago

Verified dbeadf16

fix step in adam (#1823)

szhengac committed 3 years ago

Verified de887185

Update partition_parameters.py (#1943)

manuelciosici committed 3 years ago

Verified ae43ba12

GatheredParameters - accept a tuple of params (#1941)

stas00 committed 3 years ago

Verified 73c0798b

jeffra committed 3 years ago

ff908ed7

Improve z3 trace management (#1916)

tjruwase committed 3 years ago

Verified 673cb608

Fix time error (#1934)

Codle committed 3 years ago

Verified a3b90030

[ZeRO-3] Rename confusing log message (#1932)

jeffra committed 3 years ago

Verified a8d26d6a

spell err (#1929)

kisseternity committed 3 years ago

Verified 89e37ef3

bf16 inference (#1917)

tjruwase committed 3 years ago

Verified af58f63d

Enable DeepSpeed inference on ROCm (#1922)

rraminen committed 3 years ago

Verified 96c8bf32

Older