Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

T5rpe #141

Merged
merged 7 commits into from
Feb 28, 2021
Merged

T5rpe #141

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
1967e20
Transfer of rpe from https://github.com/EleutherAI/megatron-3d/issues/9.
MicPie Feb 19, 2021
d54a195
Added Michael Pieler (MicPie) to the list of authors
StellaAthena Feb 19, 2021
5b9792f
Update to work with args 1.
MicPie Feb 21, 2021
065f138
Update to work with args 2.
MicPie Feb 21, 2021
16184ab
Cleanup.
MicPie Feb 21, 2021
2507fea
Cleanup 2.
MicPie Feb 21, 2021
3685021
Merge branch 'main' into t5rpe
sdtblck Feb 27, 2021
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
Prev Previous commit
Next Next commit
Update to work with args 1.
  • Loading branch information
@MicPie
MicPie committed Feb 21, 2021
commit 5b9792f916c44911a2e9e24bbe8e8b0037b7ad7b
9 changes: 7 additions & 2 deletions megatron/model/transformer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -586,11 +586,16 @@ class ParallelTransformer(MegatronModule):
"""Transformer class."""

def __init__(self, attention_mask_func,
init_method, output_layer_init_method,
rpe=True, rpe_causal=False, rpe_num_buckets=32, rpe_max_distance=128):
init_method, output_layer_init_method):
# rpe=True, rpe_causal=False, rpe_num_buckets=32, rpe_max_distance=128):
super(ParallelTransformer, self).__init__()
args = get_args()

self.rpe = args.rpe # True
self.rpe_causal = args.rpe_causal # False
self.rpe_num_buckets =.args.rpe_num_buckets # 32
self.rpe_max_distance = args.rpe_max_distance # 128

# Store activation checkpoiting flag.
self.checkpoint_activations = args.checkpoint_activations
self.checkpoint_num_layers = args.checkpoint_num_layers
Expand Down