Skip to content

fix curriculum learning eval mode #139

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

Merged
conglongli merged 2 commits into from
Oct 25, 2021
Merged

fix curriculum learning eval mode #139

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4c4e17a
handle None case
conglongli Oct 25, 2021
fdd8d89
Merge branch 'master' into cl_patch
conglongli Oct 25, 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
6 changes: 5 additions & 1 deletion Megatron-LM-v1.1.5-ZeRO3/megatron/model/gpt2_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,8 +55,8 @@ def __init__(self, num_tokentypes=0, parallel_output=True):
def forward(self, input_ids, position_ids, attention_mask, labels=None,
tokentype_ids=None, layer_past=None, get_key_value=False,
forward_method_parallel_output=None, curriculum_seqlen=None):
args = get_args()
if curriculum_seqlen is not None:
args = get_args()
args.curriculum_seqlen = curriculum_seqlen
if curriculum_seqlen < input_ids.size()[1]:
# seqlen-based curriculum learning
Expand All @@ -67,6 +67,10 @@ def forward(self, input_ids, position_ids, attention_mask, labels=None,

# attention_mask has size [1, 1, seqlen, seqlen]
attention_mask = attention_mask[:, :, :curriculum_seqlen, :curriculum_seqlen].contiguous()
else:
if args.curriculum_learning:
# If got a None input, need to reset curriculum_seqlen on user side
args.curriculum_seqlen = args.seq_length

# Language model.
lm_output = self.language_model(input_ids,
Expand Down