Skip to content

fix: don't use kernel layernorm on Blackwell architecture to avoid "no kernel image" error #3343

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
AdamPalaxo wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: don't use kernel layernorm on Blackwell architecture to avoid "no kernel image" error #3343

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
7 changes: 5 additions & 2 deletions server/text_generation_server/layers/layernorm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,9 +36,12 @@ def load_layer_norm_no_bias(cls, prefix, weights, eps):
if SYSTEM == "cuda":
import dropout_layer_norm

major, _ = torch.cuda.get_device_capability()
is_blackwell = major > 9

class FastLayerNorm(nn.LayerNorm):
def forward(self, hidden_states, residual=None):
if hidden_states.shape[-1] > 8192:
if hidden_states.shape[-1] > 8192 or is_blackwell:
if residual is not None:
hidden_states += residual
residual = hidden_states
Expand Down Expand Up @@ -142,7 +145,7 @@ def forward(self, hidden_states, residual=None):
self.variance_epsilon,
)
return out, residual
elif hidden_states.shape[-1] > 8192:
elif hidden_states.shape[-1] > 8192 or is_blackwell:
if residual is not None:
hidden_states += residual
residual = hidden_states
Expand Down