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

[Bug] fix preference_collate_fn attn_mask #859

Merged
merged 1 commit into from
Jul 22, 2024
Merged

[Bug] fix preference_collate_fn attn_mask #859

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
15 changes: 8 additions & 7 deletions xtuner/dataset/collate_fns/preference_collate_fn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,14 +58,14 @@ def preference_collate_fn(instances: Sequence[Dict],
labels = torch.stack(labels)

if use_varlen_attn:
attention_mask = torch.ones_like(input_ids).bool()
attention_mask = None
position_ids = torch.stack(position_ids, dim=0)
else:
# Some tokenizers have the same eos token and pad token, so input_ids
# cannot be masked directly based on the pad token id.
attention_mask = torch.zeros_like(input_ids).bool()
for i in ori_length:
attention_mask[:i] = True
for i, length in enumerate(ori_length):
attention_mask[i, :length] = True

bs, seq_len = input_ids.shape
position_ids = torch.arange(seq_len).unsqueeze(0).long().repeat(bs, 1)
Expand All @@ -74,11 +74,12 @@ def preference_collate_fn(instances: Sequence[Dict],
input_ids = pad_for_sequence_parallel(input_ids, pad_index)
labels = pad_for_sequence_parallel(labels, IGNORE_INDEX)
position_ids = pad_for_sequence_parallel(position_ids, 0)
# We use attention_mask to distinguish `input_ids` from
# (sequence parallel) pad tokens in `get_var_len_atten_logps` method of
# class `DPO` and `ORPO`
attention_mask = pad_for_sequence_parallel(attention_mask, 0)
if attention_mask is not None:
attention_mask = pad_for_sequence_parallel(attention_mask, 0)
if use_varlen_attn:
# We use attention_mask to distinguish `input_ids` from
# (sequence parallel) pad tokens in `get_var_len_atten_logps`
# method of class `DPO` and `ORPO`
(cumulative_len, attention_mask
) = pad_cumulative_len_for_sequence_parallel(cumulative_len)

Expand Down
Loading