- 27 Jan, 2022 2 commits
-
-
Stas Bekman authored
-
Stas Bekman authored
The paper has this info, so proposing to copy it next to the table. Otherwise it's hard to guess whether you used 40GB A100s or 80GB ones (and secondary, n_gpus per node). Thank you!
-
- 05 Jan, 2022 1 commit
-
-
Jared Casper authored
added flag to control deallocation of pipeline outputs See merge request ADLR/megatron-lm!377
-
- 04 Jan, 2022 3 commits
-
-
Lawrence McAfee authored
-
Jared Casper authored
Generic fix to T5 pipeline parallelism bug. See merge request ADLR/megatron-lm!376
-
Vijay Korthikanti authored
-
- 31 Dec, 2021 1 commit
-
-
Vijay Korthikanti authored
-
- 30 Dec, 2021 2 commits
-
-
Vijay Korthikanti authored
-
Vijay Korthikanti authored
-
- 21 Dec, 2021 4 commits
-
-
Jared Casper authored
pipeline_fixes See merge request ADLR/megatron-lm!372
-
Vijay Korthikanti authored
-
Jared Casper authored
Deallocate pipeline stage output tensors after sending See merge request ADLR/megatron-lm!370
-
Jared Casper authored
Multistage prompting/main multistage See merge request ADLR/megatron-lm!371
-
- 17 Dec, 2021 2 commits
-
-
Vijay Korthikanti authored
-
Vijay Korthikanti authored
-
- 14 Dec, 2021 4 commits
- 11 Dec, 2021 1 commit
-
-
zihanl authored
-
- 10 Dec, 2021 13 commits
-
-
Lawrence McAfee authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
zihanl authored
-
- 09 Dec, 2021 7 commits