Skip to content

fix(qwen3.5): average TP-replicated linear-attention gradients across TP #2451

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
gavinqii wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix(qwen3.5): average TP-replicated linear-attention gradients across TP #2451

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e6ae346
fix(qwen3.5): average TP-replicated linear-attention gradients across TP
gavinqii Oct 8, 2026
5dcf67c
Merge branch 'main' into fix/qwen3.5-tp-replicated-grad-sync
gavinqii Oct 10, 2026
7f6a043
Merge branch 'main' into fix/qwen3.5-tp-replicated-grad-sync
gavinqii Oct 12, 2026
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: 7 additions & 0 deletions slime_plugins/models/qwen3_5.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -186,6 +186,13 @@ def __init__(

self.input_layernorm = RMSNorm(self.hf_config.hidden_size, eps=self.hf_config.rms_norm_eps)

# linear_attn and input_layernorm are replicated, not sharded, across TP ranks.
# Average their gradients across TP so the replicas stay identical; without this,
# nondeterministic kernels let them drift apart a little more every step.
for module in (self.linear_attn, self.input_layernorm):
for param in module.parameters():
param.average_gradients_across_tp_domain = True

def hf_forward(self, hidden_states, packed_seq_params):
hidden_states = self.input_layernorm(hidden_states)
hidden_states = self.linear_attn(
Expand Down
Loading