Skip to content

Add apply_router_weight_on_input to all FusedMoEMethodBase classes #5

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
merged 4 commits into from
Apr 6, 2025
Merged

Add apply_router_weight_on_input to all FusedMoEMethodBase classes #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
59f8e46
Add missing apply_router_weight_on_input arg to all FusedMoEMethodBas…
sarckk Apr 5, 2025
10df398
Make linter happy
sarckk Apr 5, 2025
675b3c1
More lint fixes
sarckk Apr 6, 2025
72e87dc
Revert "More lint fixes"
sarckk Apr 6, 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
2 changes: 2 additions & 0 deletions vllm/model_executor/layers/fused_moe/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,8 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
) -> torch.Tensor:
raise NotImplementedError

Expand Down
5 changes: 5 additions & 0 deletions vllm/model_executor/layers/quantization/awq_marlin.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -469,13 +469,18 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
) -> torch.Tensor:
assert activation == "silu", "Only SiLU activation is supported."
if expert_map is not None:
raise NotImplementedError(
"Expert Parallelism is not supported for "
"fused Marlin MoE method.")
if apply_router_weight_on_input:
raise NotImplementedError(
"Apply router weight on input is not supported for"
"fused Marlin MoE method.")

topk_weights, topk_ids = FusedMoE.select_experts(
hidden_states=x,
Expand Down
27 changes: 15 additions & 12 deletions vllm/model_executor/layers/quantization/experts_int8.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,6 +113,7 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
) -> torch.Tensor:
from vllm.model_executor.layers.fused_moe import fused_experts
Expand All @@ -129,18 +130,20 @@ def apply(
scoring_func=scoring_func,
e_score_correction_bias=e_score_correction_bias)

return fused_experts(x,
layer.w13_weight,
layer.w2_weight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
activation=activation,
use_int8_w8a16=True,
global_num_experts=global_num_experts,
expert_map=expert_map,
w1_scale=layer.w13_scale,
w2_scale=layer.w2_scale)
return fused_experts(
x,
layer.w13_weight,
layer.w2_weight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
activation=activation,
use_int8_w8a16=True,
global_num_experts=global_num_experts,
apply_router_weight_on_input=apply_router_weight_on_input,
expert_map=expert_map,
w1_scale=layer.w13_scale,
w2_scale=layer.w2_scale)

@staticmethod
def quantizing_weight_loader(layer, weight_loader):
Expand Down
6 changes: 6 additions & 0 deletions vllm/model_executor/layers/quantization/gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -338,9 +338,15 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
):
assert activation == "silu", "Only SiLU activation is supported."
if apply_router_weight_on_input:
raise NotImplementedError(
"Apply router weight on input is not supported for"
"fused GGUF MoE method.")

topk_weights, topk_ids = FusedMoE.select_experts(
hidden_states=x,
router_logits=router_logits,
Expand Down
5 changes: 5 additions & 0 deletions vllm/model_executor/layers/quantization/gptq_marlin.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -592,9 +592,14 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
) -> torch.Tensor:
assert activation == "silu", "Only SiLU activation is supported."
if apply_router_weight_on_input is not None:
raise NotImplementedError(
"Apply router weight on input is not supported for"
"fused Marlin MoE method.")

# The input must currently be float16
orig_dtype = x.dtype
Expand Down
33 changes: 18 additions & 15 deletions vllm/model_executor/layers/quantization/moe_wna16.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -293,6 +293,7 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
Copy link
Collaborator

@luccafong luccafong Apr 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this should be above "activation", if you run "pre-commit run --show-diff-on-failure --color=always --all-files --hook-stage manual" there would be a failure that it does not compatible with the base signature

activation: str = "silu",
) -> torch.Tensor:
from vllm.model_executor.layers.fused_moe import fused_experts
Expand All @@ -312,21 +313,23 @@ def apply(
weight_bits = self.quant_config.weight_bits
has_zp = self.quant_config.has_zp

return fused_experts(x,
layer.w13_qweight,
layer.w2_qweight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
use_int4_w4a16=weight_bits == 4,
use_int8_w8a16=weight_bits == 8,
global_num_experts=global_num_experts,
expert_map=expert_map,
w1_scale=layer.w13_scales,
w2_scale=layer.w2_scales,
w1_zp=layer.w13_qzeros if has_zp else None,
w2_zp=layer.w2_qzeros if has_zp else None,
block_shape=[0, layer.group_size])
return fused_experts(
x,
layer.w13_qweight,
layer.w2_qweight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
use_int4_w4a16=weight_bits == 4,
use_int8_w8a16=weight_bits == 8,
global_num_experts=global_num_experts,
apply_router_weight_on_input=apply_router_weight_on_input,
expert_map=expert_map,
w1_scale=layer.w13_scales,
w2_scale=layer.w2_scales,
w1_zp=layer.w13_qzeros if has_zp else None,
w2_zp=layer.w2_qzeros if has_zp else None,
block_shape=[0, layer.group_size])

@staticmethod
def get_weight_loader(layer, weight_loader):
Expand Down
30 changes: 17 additions & 13 deletions vllm/model_executor/layers/quantization/quark/quark_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,6 +202,8 @@ def apply(
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None,
apply_router_weight_on_input: bool = False,
activation: str = "silu",
) -> torch.Tensor:
from vllm.model_executor.layers.fused_moe import fused_experts

Expand All @@ -217,16 +219,18 @@ def apply(
scoring_func=scoring_func,
e_score_correction_bias=e_score_correction_bias)

return fused_experts(x,
Copy link
Collaborator

@luccafong luccafong Apr 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: this seems no need to change, the format previously is following vllm format style

Copy link
Collaborator Author

@sarckk sarckk Apr 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

formatted changed it because we are passing apply_router_weight_on_input=apply_router_weight_on_input now

layer.w13_weight,
layer.w2_weight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
use_fp8_w8a8=True,
global_num_experts=global_num_experts,
expert_map=expert_map,
w1_scale=layer.w13_weight_scale,
w2_scale=layer.w2_weight_scale,
a1_scale=layer.w13_input_scale,
a2_scale=layer.w2_input_scale)
return fused_experts(
x,
layer.w13_weight,
layer.w2_weight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
use_fp8_w8a8=True,
global_num_experts=global_num_experts,
apply_router_weight_on_input=apply_router_weight_on_input,
expert_map=expert_map,
w1_scale=layer.w13_weight_scale,
w2_scale=layer.w2_weight_scale,
a1_scale=layer.w13_input_scale,
a2_scale=layer.w2_input_scale)