- 12 Jan, 2021 1 commit
-
-
Mohammad Shoeybi authored
-
- 07 Jan, 2021 2 commits
-
-
Mohammad Shoeybi authored
Change some arguments to default to on. See merge request ADLR/megatron-lm!203
-
Jared Casper authored
-
- 06 Jan, 2021 4 commits
-
-
Jared Casper authored
Optimizer refactor See merge request ADLR/megatron-lm!197
-
Jared Casper authored
added backward compatibility to be able to load checkpoints See merge request ADLR/megatron-lm!200
-
mohammad authored
-
mohammad authored
-
- 05 Jan, 2021 13 commits
-
-
mohammad authored
-
mohammad authored
-
Mohammad Shoeybi authored
Only create task heads on last pipeline stage. See merge request ADLR/megatron-lm!202
-
Mohammad Shoeybi authored
Use new api to get loss scale when finetuning. See merge request ADLR/megatron-lm!201
-
Jared Casper authored
-
Jared Casper authored
-
Jared Casper authored
Optimize refactor reload model params See merge request ADLR/megatron-lm!199
-
Jared Casper authored
Use batched send and recv instead of torch.distributed.ring_exchange() See merge request ADLR/megatron-lm!198
-
mohammad authored
-
mohammad authored
-
mohammad authored
-
mohammad authored
-
mohammad authored
-
- 04 Jan, 2021 1 commit
-
-
Deepak Narayanan authored
-
- 31 Dec, 2020 1 commit
-
-
mohammad authored
-
- 30 Dec, 2020 7 commits
- 28 Dec, 2020 1 commit
-
-
Mohammad authored
-
- 27 Dec, 2020 2 commits
- 25 Dec, 2020 7 commits
- 23 Dec, 2020 1 commit
-
-
Mohammad Shoeybi authored
Run evaluation pass before checkpoint is saved See merge request ADLR/megatron-lm!196
-
