Skip to content

[Bugfix] Disable sequence parallelism for Dots3 NOTE #52172

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
youkaichao merged 1 commit into from
Aug 13, 2026
Merged

[Bugfix] Disable sequence parallelism for Dots3 NOTE #52172

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: 3 additions & 5 deletions vllm/models/dots3_note/nvidia/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -497,6 +497,7 @@ def __init__(
layer_idx = int(prefix.split(sep=".")[-1])
self.layer_idx = layer_idx
self.use_mha = False
self.use_sequence_parallel = False
attention_cls = (
Dots3NoteSlidingAttention
if config.layer_types[layer_idx] == "sliding_attention"
Expand Down Expand Up @@ -536,11 +537,7 @@ def __init__(
prefix=f"{prefix}.mlp",
reduce_results=False,
)
self.use_sequence_parallel_moe = (
parallel_config.use_sequence_parallel_moe
and parallel_config.pipeline_parallel_size == 1
and isinstance(self.mlp, DeepseekV2MoE)
)
self.use_sequence_parallel_moe = False
self.tp_size = parallel_config.tensor_parallel_size
self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.post_attention_layernorm = RMSNorm(
Expand All @@ -559,6 +556,7 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = "") -> None:
self.config = config
self.device = current_platform.device_type
self.vocab_size = config.vocab_size
self.use_sequence_parallel = False
self.is_v32 = True
self._weight_block_size = getattr(quant_config, "weight_block_size", None)
topk_indices_buffer = torch.empty(
Expand Down
Loading