Skip to content

[NVIDIA] Enable trtllm fp4 moe for qwen #13556

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

Closed
kaixih wants to merge 2 commits into from
Closed

[NVIDIA] Enable trtllm fp4 moe for qwen #13556

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
58675c7
Enable trtllm fp4 moe for qwen
kaixih Nov 19, 2025
a7e3808
Lint
kaixih Nov 19, 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
24 changes: 19 additions & 5 deletions python/sglang/srt/layers/moe/fused_moe_triton/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1083,8 +1083,6 @@ def forward(self, hidden_states: torch.Tensor, topk_output: TopKOutput):

hs_fp4, hs_scale_linear = self._quantize_hidden_states_fp4(hidden_states)

router_logits = router_logits.to(torch.float32)

with use_symmetric_memory(
get_tp_group(), disabled=not is_allocation_symmetric()
):
Expand All @@ -1097,9 +1095,25 @@ def forward(self, hidden_states: torch.Tensor, topk_output: TopKOutput):
symm_output = torch.empty(
num_tokens, hidden_size, dtype=torch.bfloat16, device=hs_fp4.device
)

# Some dtype or value requirements for the flashinfer trtllm kernels.
router_logits = (
router_logits.to(torch.float32)
if self.routing_method_type == RoutingMethodType.DeepSeekV3
else router_logits
)
correction_bias = (
None
if topk_config.correction_bias is None
else topk_config.correction_bias.to(hidden_states.dtype)
)
routed_scaling_factor = self.moe_runner_config.routed_scaling_factor
if routed_scaling_factor is None:
routed_scaling_factor = 1.0

result = trtllm_fp4_block_scale_moe(
routing_logits=router_logits,
routing_bias=topk_config.correction_bias.to(hidden_states.dtype),
routing_bias=correction_bias,
hidden_states=hs_fp4,
hidden_states_scale=hs_scale_linear.view(torch.float8_e4m3fn).flatten(),
gemm1_weights=self.gemm1_weights_fp4_shuffled.data,
Expand All @@ -1125,9 +1139,9 @@ def forward(self, hidden_states: torch.Tensor, topk_output: TopKOutput):
intermediate_size=self.intermediate_size_per_partition,
local_expert_offset=self.moe_ep_rank * self.num_local_experts,
local_num_experts=self.num_local_experts,
routed_scaling_factor=self.moe_runner_config.routed_scaling_factor,
routed_scaling_factor=routed_scaling_factor,
tile_tokens_dim=None,
routing_method_type=RoutingMethodType.DeepSeekV3,
routing_method_type=self.routing_method_type,
do_finalize=True,
output=symm_output,
)[0]
Expand Down
Loading