Skip to content
This repository was archived by the owner on Jan 15, 2026. It is now read-only.

Setup mcore T5 for mT5 model #386

Open
huvunvidia wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Setup mcore T5 for mT5 model #386

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions launcher_scripts/conf/training/mt5/170m.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,8 +53,8 @@ exp_manager:
buffer_size: 5

model:
mcore_t5: False
transformer_engine: False
mcore_t5: True
transformer_engine: True

# model parallelism
micro_batch_size: 64
Expand Down Expand Up @@ -215,4 +215,4 @@ model:
favor_longer_ngrams: False
respect_document_boundaries: True # If true, a single training exampl cannot cross document boundaries, increasing the fraction of <pad> tokens within a batch.
index_mapping_dir: null # path to save index mapping .npy files, by default will save in the same location as data_prefix
data_prefix: null # Should be weight path weight path... for a blended dataset. If null will automatically blend all language files in mC4_dir.
data_prefix: null # Should be weight path weight path... for a blended dataset. If null will automatically blend all language files in mC4_dir.