bigscience-workshop/Megatron-DeepSpeed

Pull Requests Commits

thomasw21 committed 4 years ago

c3ab5b95

thomasw21 committed 4 years ago

d1605898

thomasw21 committed 4 years ago

9a5bf96d

thomasw21 committed 4 years ago

32749863

thomasw21 committed 4 years ago

ea221a88

thomasw21 committed 4 years ago

c53622a9

thomasw21 committed 4 years ago

014b8b82

thomasw21 committed 4 years ago

0e34e8d1

thomasw21 committed 4 years ago

ddc6a61a

thomasw21 committed 4 years ago

688d06e4

Work into moving packing logic into a dataset

thomasw21 committed 4 years ago

273d420b

Make MTF dataloader an infinite dataloader

thomasw21 committed 4 years ago

84007bc2

MTF dataset and packing (#293)

thomasw21 committed 4 years ago

Verified c5b88fb9

Use bloom instead

thomasw21 committed 4 years ago

8b67bd98

thomasw21 committed 4 years ago

da9f3160

Maybe nothing is wrong

thomasw21 committed 4 years ago

7eaced45

Test something out

thomasw21 committed 4 years ago

65b4ea28

Test something out

thomasw21 committed 4 years ago

723a5b39

Test something out

thomasw21 committed 4 years ago

8b31e9ca

Test something out

thomasw21 committed 4 years ago

964a275f

Test something out

thomasw21 committed 4 years ago

0e1ef5dc

Test something out

thomasw21 committed 4 years ago

292620c4

Test something out

thomasw21 committed 4 years ago

5746641e

Test something out

thomasw21 committed 4 years ago

3a846a0a

Test something out

thomasw21 committed 4 years ago

ed6131aa

Test something out

thomasw21 committed 4 years ago

c34f1073

Head specific mask is not implemented

thomasw21 committed 4 years ago

75f99ef7

Add a test checking that ScaledMaskedSoftmax custom kernel does what we expect

thomasw21 committed 4 years ago

910f93b9

Revert "Probably should be in this order:"

thomasw21 committed 4 years ago

f6d717b4

Probably should be in this order:

thomasw21 committed 4 years ago

93cde870

Newer Older