Skip to content

[shardformer]t5 flash attention fix #4239

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
CjhHa1 merged 11 commits into from
Jul 14, 2023
Merged

[shardformer]t5 flash attention fix #4239

Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
cb9d34b
[shardformer] opt support flash attention
flybird11111 Jul 4, 2023
ac179b5
[shardformer] opt support flash attention
flybird11111 Jul 4, 2023
00c4a82
[shardformer] opt support flash attention
flybird11111 Jul 4, 2023
27526d9
[shardformer] opt support flash attention
flybird11111 Jul 4, 2023
70535bd
[shardformer] move to modeling
flybird11111 Jul 5, 2023
794dd86
[shardformer] move to modeling
flybird11111 Jul 5, 2023
60f4a36
Merge branch 'hpcaitech:feature/flash-attention-shardformer' into fea…
flybird11111 Jul 6, 2023
4dd32cc
Merge branch 'hpcaitech:feature/flash-attention-shardformer' into fea…
flybird11111 Jul 6, 2023
a0f97e1
Merge branch 'hpcaitech:feature/flash-attention-shardformer' into fea…
flybird11111 Jul 13, 2023
3fecf78
Merge branch 'hpcaitech:feature/flash-attention-shardformer' into fea…
flybird11111 Jul 14, 2023
db6acd6
[shardformer] t5 flash attention fix
flybird11111 Jul 14, 2023
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
3 changes: 0 additions & 3 deletions colossalai/shardformer/modeling/t5.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,9 +30,6 @@ def t5_flash_attention_forward(
# Input is (batch_size, seq_length, dim)
# Mask is (batch_size, key_length) (non-causal) or (batch_size, key_length, key_length)
# past_key_value[0] is (batch_size, n_heads, q_len - 1, dim_per_head)
import pathlib
pathlib.Path("/home/lcjmy/code/personal/ColossalAI/colossalai/shardformer/modeling/mask.txt").write_text(
str(mask) + str(mask.shape))
batch_size, seq_length = hidden_states.shape[:2]

real_seq_length = seq_length
Expand Down