Skip to content

qwen4_exp: build o_proj row-parallel, as every other family does #429

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
gberasmus87 wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −2
Open

qwen4_exp: build o_proj row-parallel, as every other family does #429

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
8 changes: 6 additions & 2 deletions python/freetoken/models/qwen4_exp/attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@

import torch
from freetoken.core import get_global_ctx
from freetoken.layers import BaseOP, GemmaPlusOneRMSNorm, LinearColParallelMerged, LinearReplicated
from freetoken.layers import BaseOP, GemmaPlusOneRMSNorm, LinearColParallelMerged, LinearOProj, LinearReplicated
from freetoken.layers.rotary import get_rope
from freetoken.utils import nvtx_annotate

Expand Down Expand Up @@ -128,7 +128,11 @@ def __init__(self, config: ModelConfig, layer_id: int, *, prefix: str = "") -> N
config.hidden_size, self._qkv_split, has_bias=False,
quant_config=config.quant, prefix=f"{prefix}.qkv_proj",
)
self.o_proj = LinearReplicated(
# Row-parallel, as every other family builds o_proj: qkv_proj is column-parallel, so a
# rank's attention output is its local head slice and the partial sums need an
# all-reduce. At TP=1 this is LinearReplicated exactly -- div_even(x, 1) == x and the
# all-reduce is skipped -- so it is a no-op today and correct when #385 lands.
self.o_proj = LinearOProj(
self.qo_attn_dim, config.hidden_size, has_bias=False,
quant_config=config.quant, prefix=f"{prefix}.o_proj",
)
Expand Down