Skip to content

[bugs] fix for casual mask #2868

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
danielhanchen merged 4 commits into from
Jul 9, 2025
Merged

[bugs] fix for casual mask #2868

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
812a97f
fix for casual mask
leizhenyuan Jul 3, 2025
05ea2a9
use un_casual in sdpa
leizhenyuan Jul 3, 2025
a9690fb
add missing mask
leizhenyuan Jul 4, 2025
53f0120
fix for type
leizhenyuan Jul 4, 2025
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
18 changes: 16 additions & 2 deletions unsloth/models/llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -318,6 +318,13 @@ def LlamaAttention_fast_forward_inference(
# Knn, Vnn = Knn, Vnn
# pass

# when qlen==vlen and attn_mask is None, we should use causal attention
Q_len = Qn.shape[-2]
K_len = Knn.shape[-2]
if attention_mask is None and Q_len == K_len:
is_causal = True
else:
is_causal = False
Copy link
Copy Markdown

@Oscilloscope98 Oscilloscope98 Jul 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks like the usage of is_causal argument is also forgotten here: https://github.com/unslothai/unsloth/pull/2868/files#diff-a45b72bb533eda979990bd79cde5fe9c9fde424779a4f1fc1195b75853d93b45R337-R340

# Attention
if bsz == 1:
Qn *= self.scalar # See https://github.com/ggerganov/llama.cpp/issues/7805#issuecomment-2153349963
Expand Down Expand Up @@ -518,11 +525,18 @@ def LlamaAttention_fast_forward(
V = V.transpose(1, 2)
A = flash_attn_func(Q, K, V, causal = True)
else:
# when qlen==vlen and attn_mask is None, we should use causal attention
Q_len = Q.shape[-2]
K_len = K.shape[-2]
if attention_mask is None and Q_len == K_len:
is_causal = True
else:
is_causal = False
Comment thread
leizhenyuan marked this conversation as resolved.
# Grouped query attention
if SDPA_HAS_GQA:
# Needs (batch_size, n_heads, seq_len, head_dim)
# is_casual and attention_mask must not be both set!
A = scaled_dot_product_attention(Q, K, V, attn_mask = attention_mask, is_causal = False, enable_gqa = n_groups != 1)
A = scaled_dot_product_attention(Q, K, V, attn_mask = attention_mask, is_causal = is_causal, enable_gqa = n_groups != 1)
# Go back to (batch_size, seq_len, n_heads, head_dim)
A = A.transpose(1, 2)#.contiguous()
else:
Expand All @@ -537,7 +551,7 @@ def LlamaAttention_fast_forward(
Q, K, V = Q.contiguous(), K.contiguous(), V.contiguous()
# Needs (batch_size, n_heads, seq_len, head_dim)
# is_casual and attention_mask must not be both set!
A = scaled_dot_product_attention(Q, K, V, attn_mask = attention_mask, is_causal = False)
A = scaled_dot_product_attention(Q, K, V, attn_mask = attention_mask, is_causal = is_causal)
# Go back to (batch_size, seq_len, n_heads, head_dim)
A = A.transpose(1, 2).contiguous()
pass
Expand Down