Skip to content

[tx] (Not for merging) Test cudnn backend for attention implementation #777

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

Open
pcmoritz wants to merge 2 commits into
base: main
Choose a base branch
from
Open

[tx] (Not for merging) Test cudnn backend for attention implementation #777

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e01e9b5
[tx] Try cudnn attention implementation
pcmoritz Dec 12, 2025
b6a89fa
update
pcmoritz Dec 12, 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
10 changes: 9 additions & 1 deletion skyrl-tx/tx/models/qwen3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -122,13 +122,21 @@ def __call__(
updated_cache = (k, v)

# Attention (causal only during prefill, GQA handled natively by dot_product_attention)
# Expand mask to [B, 1, Q_len, KV_len] for cuDNN compatibility (no broadcasting)
Q_len = q.shape[1]
KV_len = k.shape[1]
mask = jnp.broadcast_to(
attention_mask[:, None, None, :].astype(bool),
(B, 1, Q_len, KV_len)
)
attn_output = jax.nn.dot_product_attention(
q,
k,
v,
scale=1.0 / self.head_dim**0.5,
mask=attention_mask[:, None, None, :].astype(bool),
mask=mask,
is_causal=kv_cache is None,
implementation="cudnn",
)
Comment on lines 132 to 140
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 12, 2025

Choose a reason for hiding this comment

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

critical

According to the JAX documentation for jax.nn.dot_product_attention with implementation="cudnn", the deterministic parameter must be set to True. Without this, the call might fail or not use the cuDNN kernel as intended.

Suggested change
attn_output = jax.nn.dot_product_attention(
q,
k,
v,
scale=1.0 / self.head_dim**0.5,
mask=attention_mask[:, None, None, :].astype(bool),
mask=mask,
is_causal=kv_cache is None,
implementation="cudnn",
)
attn_output = jax.nn.dot_product_attention(
q,
k,
v,
scale=1.0 / self.head_dim**0.5,
mask=mask,
is_causal=kv_cache is None,
implementation="cudnn",
deterministic=True,
)


output = attn_output.reshape(B, T, self.num_heads * self.head_dim)
Expand Down
Loading