- 07 Jul, 2021 1 commit
-
-
rprenger authored
Getting rid of profiling and adding route for index (it sees the root of the project as the megatron subdirectory for some reason)
-
- 02 Jul, 2021 1 commit
-
-
rprenger authored
-
- 30 Jun, 2021 3 commits
- 23 Jun, 2021 2 commits
- 01 Jun, 2021 2 commits
-
-
Jared Casper authored
Finetuning retriever (ICT+DPR) See merge request ADLR/megatron-lm!277
-
Jared Casper authored
vit pipeline fixes See merge request ADLR/megatron-lm!276
-
- 25 May, 2021 1 commit
-
-
Mostofa Patwary authored
-
- 20 May, 2021 2 commits
-
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
- 19 May, 2021 9 commits
-
-
Mohammad Shoeybi authored
Add temporary assert to finetuning until it can be fixed. See merge request ADLR/megatron-lm!278
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
Jared Casper authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
- 18 May, 2021 4 commits
-
-
Mostofa Patwary authored
-
Vijay Korthikanti authored
-
Mostofa Patwary authored
-
Mostofa Patwary authored
-
- 15 May, 2021 4 commits
-
-
Jared Casper authored
Bringing github main and internal main back together. See merge request ADLR/megatron-lm!275
-
Jared Casper authored
-
Jared Casper authored
One more quick fix. See merge request ADLR/megatron-lm!274
-
Jared Casper authored
-
- 14 May, 2021 11 commits
-
-
Jared Casper authored
Quick fix of copy/paste error. See merge request ADLR/megatron-lm!273
-
Jared Casper authored
-
Mohammad Shoeybi authored
Add arg needed for tokenizer. See merge request ADLR/megatron-lm!272
-
Jared Casper authored
-
Jared Casper authored
Ensure numpy random seed is within range. See merge request ADLR/megatron-lm!271
-
Jared Casper authored
-
Mohammad Shoeybi authored
Update arguments checks. See merge request ADLR/megatron-lm!270
-
Jared Casper authored
Adding T5 to docs and a bit of cleanup. See merge request ADLR/megatron-lm!269
-
Jared Casper authored
-
Jared Casper authored
hidden_size % attention_heads == 0 is handled above when dealing with kv_channels. Adding check for decoder sequence length.
-
Jared Casper authored
fixed compatiblity with v0 checkpoints See merge request ADLR/megatron-lm!268
-