Skip to content

vllm.compilation.collective_fusion

ALLREDUCE_OP module-attribute 

ALLREDUCE_OP = default

FP8_DTYPE module-attribute 

FP8_DTYPE = fp8_dtype()

MiB module-attribute 

MiB = 1024 * 1024

RMS_ADD_OP module-attribute 

RMS_ADD_OP = default

RMS_OP module-attribute 

RMS_OP = default

STATIC_FP4_QUANT_OP module-attribute 

STATIC_FP4_QUANT_OP = default

STATIC_FP8_QUANT_OP module-attribute 

STATIC_FP8_QUANT_OP = default

_DEFAULT_FI_MAX_SIZE module-attribute 

_DEFAULT_FI_MAX_SIZE = MiB // 2

_FI_MAX_SIZES module-attribute 

_FI_MAX_SIZES = {
    2: 64 * MiB,
    4: MiB,
    6: MiB // 2,
    8: MiB // 2,
}

_FI_WORKSPACE_TENSOR module-attribute 

_FI_WORKSPACE_TENSOR = None

flashinfer_trtllm_fused_allreduce_norm module-attribute 

flashinfer_trtllm_fused_allreduce_norm = default

logger module-attribute 

logger = init_logger(__name__)

AllGatherCutlassScaledMMPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
class AllGatherCutlassScaledMMPattern(BasePattern):

    def get_inputs(self):
        x = torch.empty([8, 16], device=self.device, dtype=FP8_DTYPE)
        weight = torch.empty([16, 16], device=self.device,
                             dtype=FP8_DTYPE).contiguous().transpose(0, 1)

        s1 = x.shape[0] * self.tp_size

        scale_a = torch.empty([s1, 1], device=self.device, dtype=torch.float32)
        scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

        s2 = weight.shape[1]
        output = torch.empty([s1, s2], device=self.device, dtype=self.dtype)

        return [x, weight, scale_a, scale_b, output]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(
            x: torch.Tensor,
            weight: torch.Tensor,
            scale_a: torch.Tensor,
            scale_b: torch.Tensor,
            output: torch.Tensor,
        ) -> torch.Tensor:
            all_gather = torch.ops.vllm.all_gather.default(
                x,
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name)

            cutlass_scaled_mm = torch.ops.higher_order.auto_functionalized(
                torch.ops._C.cutlass_scaled_mm.default,
                out=output,
                a=all_gather,
                b=weight,
                a_scales=scale_a,
                b_scales=scale_b,
                bias=None)
            return cutlass_scaled_mm[1]

        def replacement(x: torch.Tensor, weight: torch.Tensor,
                        scale_a: torch.Tensor, scale_b: torch.Tensor,
                        output: torch.Tensor) -> torch.Tensor:
            ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_scaled_matmul(  # noqa
                x,
                [weight],
                scale_a,
                [scale_b],
                gather_dim=0,
                biases=[None],
                result_scales=[None],
                out_dtypes=[self.dtype],
                use_fast_accum=[False],
                group_name=self.tp.device_group.group_name,
            )
            return mm_outputs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
289
290
291
292
293
294
295
296
297
298
299
300
301
302
def get_inputs(self):
    x = torch.empty([8, 16], device=self.device, dtype=FP8_DTYPE)
    weight = torch.empty([16, 16], device=self.device,
                         dtype=FP8_DTYPE).contiguous().transpose(0, 1)

    s1 = x.shape[0] * self.tp_size

    scale_a = torch.empty([s1, 1], device=self.device, dtype=torch.float32)
    scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

    s2 = weight.shape[1]
    output = torch.empty([s1, s2], device=self.device, dtype=self.dtype)

    return [x, weight, scale_a, scale_b, output]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
def register(self, pm_pass: PatternMatcherPass):

    def pattern(
        x: torch.Tensor,
        weight: torch.Tensor,
        scale_a: torch.Tensor,
        scale_b: torch.Tensor,
        output: torch.Tensor,
    ) -> torch.Tensor:
        all_gather = torch.ops.vllm.all_gather.default(
            x,
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name)

        cutlass_scaled_mm = torch.ops.higher_order.auto_functionalized(
            torch.ops._C.cutlass_scaled_mm.default,
            out=output,
            a=all_gather,
            b=weight,
            a_scales=scale_a,
            b_scales=scale_b,
            bias=None)
        return cutlass_scaled_mm[1]

    def replacement(x: torch.Tensor, weight: torch.Tensor,
                    scale_a: torch.Tensor, scale_b: torch.Tensor,
                    output: torch.Tensor) -> torch.Tensor:
        ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_scaled_matmul(  # noqa
            x,
            [weight],
            scale_a,
            [scale_b],
            gather_dim=0,
            biases=[None],
            result_scales=[None],
            out_dtypes=[self.dtype],
            use_fast_accum=[False],
            group_name=self.tp.device_group.group_name,
        )
        return mm_outputs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

AllGatherGEMMPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
class AllGatherGEMMPattern(BasePattern):

    def get_inputs(self):
        x = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)

        return [x, weight]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(
            x: torch.Tensor,
            weight: torch.Tensor,
        ) -> tuple[torch.Tensor, torch.Tensor]:
            all_gather = torch.ops.vllm.all_gather.default(
                x,
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name,
            )

            return torch.ops.aten.mm.default(all_gather, weight)

        def replacement(
                x: torch.Tensor,
                weight: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
            ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_matmul(
                x,
                [weight],
                gather_dim=0,
                group_name=self.tp.device_group.group_name,
            )
            return mm_outputs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
90
91
92
93
94
def get_inputs(self):
    x = torch.empty([4, 4], device=self.device, dtype=self.dtype)
    weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)

    return [x, weight]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
def register(self, pm_pass: PatternMatcherPass):

    def pattern(
        x: torch.Tensor,
        weight: torch.Tensor,
    ) -> tuple[torch.Tensor, torch.Tensor]:
        all_gather = torch.ops.vllm.all_gather.default(
            x,
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name,
        )

        return torch.ops.aten.mm.default(all_gather, weight)

    def replacement(
            x: torch.Tensor,
            weight: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
        ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_matmul(
            x,
            [weight],
            gather_dim=0,
            group_name=self.tp.device_group.group_name,
        )
        return mm_outputs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

AllGatherScaledMMPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
class AllGatherScaledMMPattern(BasePattern):

    def get_inputs(self):
        x = torch.empty([8, 16], device=self.device, dtype=FP8_DTYPE)
        weight = torch.empty([16, 16], device=self.device,
                             dtype=FP8_DTYPE).contiguous().transpose(0, 1)

        s1 = x.shape[0] * self.tp_size

        scale_a = torch.empty([s1, 1], device=self.device, dtype=torch.float32)
        scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

        return [x, weight, scale_a, scale_b]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(
            x: torch.Tensor,
            weight: torch.Tensor,
            scale_a: torch.Tensor,
            scale_b: torch.Tensor,
        ) -> torch.Tensor:
            all_gather = torch.ops.vllm.all_gather.default(
                x,
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name)

            return torch.ops.aten._scaled_mm.default(all_gather,
                                                     mat2=weight,
                                                     scale_a=scale_a,
                                                     scale_b=scale_b,
                                                     bias=None,
                                                     scale_result=None,
                                                     out_dtype=self.dtype)

        def replacement(x: torch.Tensor, weight: torch.Tensor,
                        scale_a: torch.Tensor,
                        scale_b: torch.Tensor) -> torch.Tensor:
            ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_scaled_matmul(  # noqa
                x,
                [weight],
                scale_a,
                [scale_b],
                gather_dim=0,
                biases=[None],
                result_scales=[None],
                out_dtypes=[self.dtype],
                use_fast_accum=[False],
                group_name=self.tp.device_group.group_name,
            )
            return mm_outputs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
177
178
179
180
181
182
183
184
185
186
187
def get_inputs(self):
    x = torch.empty([8, 16], device=self.device, dtype=FP8_DTYPE)
    weight = torch.empty([16, 16], device=self.device,
                         dtype=FP8_DTYPE).contiguous().transpose(0, 1)

    s1 = x.shape[0] * self.tp_size

    scale_a = torch.empty([s1, 1], device=self.device, dtype=torch.float32)
    scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

    return [x, weight, scale_a, scale_b]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
def register(self, pm_pass: PatternMatcherPass):

    def pattern(
        x: torch.Tensor,
        weight: torch.Tensor,
        scale_a: torch.Tensor,
        scale_b: torch.Tensor,
    ) -> torch.Tensor:
        all_gather = torch.ops.vllm.all_gather.default(
            x,
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name)

        return torch.ops.aten._scaled_mm.default(all_gather,
                                                 mat2=weight,
                                                 scale_a=scale_a,
                                                 scale_b=scale_b,
                                                 bias=None,
                                                 scale_result=None,
                                                 out_dtype=self.dtype)

    def replacement(x: torch.Tensor, weight: torch.Tensor,
                    scale_a: torch.Tensor,
                    scale_b: torch.Tensor) -> torch.Tensor:
        ag_output, mm_outputs = torch.ops.symm_mem.fused_all_gather_scaled_matmul(  # noqa
            x,
            [weight],
            scale_a,
            [scale_b],
            gather_dim=0,
            biases=[None],
            result_scales=[None],
            out_dtypes=[self.dtype],
            use_fast_accum=[False],
            group_name=self.tp.device_group.group_name,
        )
        return mm_outputs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusedAddRMSNormPattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (with residual) with fused flashinfer implementation. Applies to o_proj + rmsnorm after attn and mlp + rmsnorm before attn.

Source code in vllm/compilation/collective_fusion.py 
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
class AllReduceFusedAddRMSNormPattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (with residual) 
    with fused flashinfer implementation.
    Applies to o_proj + rmsnorm after attn and mlp + rmsnorm before attn.
    """

    def __init__(
        self,
        epsilon: float,
        dtype: torch.dtype,
        device: str,
        allreduce_params: FlashInferFusedAllReduceParams,
    ):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params

    def get_inputs(self):
        input = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        residual = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        return [
            residual,
            input,
            weight,
        ]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(residual: torch.Tensor, input: torch.Tensor,
                    weight: torch.Tensor):
            allreduce_output = tensor_model_parallel_all_reduce(input)
            rms = auto_functionalized(
                RMS_ADD_OP,
                input=allreduce_output,
                residual=residual,
                weight=weight,
                epsilon=self.epsilon,
            )
            # input, residual
            return rms[1], rms[2]

        def replacement(residual: torch.Tensor, input: torch.Tensor,
                        weight: torch.Tensor):
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=None,
                quant_out=None,
                scale_out=None,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNorm,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )
            # allreduce_in, residual
            return allreduce[1], allreduce[2]

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
660
661
662
663
664
665
666
667
668
669
def __init__(
    self,
    epsilon: float,
    dtype: torch.dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
671
672
673
674
675
676
677
678
679
def get_inputs(self):
    input = torch.empty([4, 4], device=self.device, dtype=self.dtype)
    residual = torch.empty([4, 4], device=self.device, dtype=self.dtype)
    weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
    return [
        residual,
        input,
        weight,
    ]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
def register(self, pm_pass: PatternMatcherPass):

    def pattern(residual: torch.Tensor, input: torch.Tensor,
                weight: torch.Tensor):
        allreduce_output = tensor_model_parallel_all_reduce(input)
        rms = auto_functionalized(
            RMS_ADD_OP,
            input=allreduce_output,
            residual=residual,
            weight=weight,
            epsilon=self.epsilon,
        )
        # input, residual
        return rms[1], rms[2]

    def replacement(residual: torch.Tensor, input: torch.Tensor,
                    weight: torch.Tensor):
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=None,
            quant_out=None,
            scale_out=None,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNorm,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )
        # allreduce_in, residual
        return allreduce[1], allreduce[2]

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusedAddRMSNormStaticQuantFP8Pattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (with residual) + static fp8 quant with fused flashinfer implementation. Applies to o_proj + rmsnorm after attn + quant and mlp + rmsnorm + quant before attn.

Source code in vllm/compilation/collective_fusion.py 
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
class AllReduceFusedAddRMSNormStaticQuantFP8Pattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (with residual)
    + static fp8 quant with fused flashinfer implementation.
    Applies to o_proj + rmsnorm after attn + quant and 
    mlp + rmsnorm + quant before attn.
    """

    def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
                 allreduce_params: FlashInferFusedAllReduceParams):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params
        self.quant_dtype = torch.float8_e4m3fn

    def register(self, pm_pass: PatternMatcherPass):

        def get_inputs():
            input = torch.empty([4, 4], device=self.device, dtype=self.dtype)

            residual = torch.empty([4, 4],
                                   device=self.device,
                                   dtype=self.dtype)
            weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
            quant_result = torch.empty([4, 4],
                                       device=self.device,
                                       dtype=self.quant_dtype)
            scale = torch.empty([1, 1],
                                device=self.device,
                                dtype=torch.float32)

            return [
                quant_result,
                residual,
                input,
                weight,
                scale,
            ]

        def pattern(
            quant_result: torch.Tensor,
            residual: torch.Tensor,
            input: torch.Tensor,
            weight: torch.Tensor,
            scale: torch.Tensor,
        ):
            allreduce_output = tensor_model_parallel_all_reduce(input)

            fused_add_rmsnorm_out_tuple = \
            auto_functionalized(
                RMS_ADD_OP,
                input=allreduce_output,
                residual=residual,
                weight=weight,
                epsilon=self.epsilon)
            quant_out_tuple = auto_functionalized(
                STATIC_FP8_QUANT_OP,
                result=quant_result,
                input=fused_add_rmsnorm_out_tuple[1],
                scale=scale)

            # quant_out, allreduce_output
            return quant_out_tuple[1], fused_add_rmsnorm_out_tuple[2]

        def replacement(quant_result: torch.Tensor, residual: torch.Tensor,
                        input: torch.Tensor, weight: torch.Tensor,
                        scale: torch.Tensor):
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=None,
                quant_out=quant_result,
                scale_out=None,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNormFP8Quant,  # we don't use norm_out afterwards
                scale_factor=scale,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )
            # # quant_out, rms_norm_residual
            return allreduce[4], allreduce[2]

        pm.register_replacement(pattern, replacement, get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

quant_dtype instance-attribute 

quant_dtype = float8_e4m3fn

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
805
806
807
808
809
810
def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
             allreduce_params: FlashInferFusedAllReduceParams):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params
    self.quant_dtype = torch.float8_e4m3fn

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
def register(self, pm_pass: PatternMatcherPass):

    def get_inputs():
        input = torch.empty([4, 4], device=self.device, dtype=self.dtype)

        residual = torch.empty([4, 4],
                               device=self.device,
                               dtype=self.dtype)
        weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        quant_result = torch.empty([4, 4],
                                   device=self.device,
                                   dtype=self.quant_dtype)
        scale = torch.empty([1, 1],
                            device=self.device,
                            dtype=torch.float32)

        return [
            quant_result,
            residual,
            input,
            weight,
            scale,
        ]

    def pattern(
        quant_result: torch.Tensor,
        residual: torch.Tensor,
        input: torch.Tensor,
        weight: torch.Tensor,
        scale: torch.Tensor,
    ):
        allreduce_output = tensor_model_parallel_all_reduce(input)

        fused_add_rmsnorm_out_tuple = \
        auto_functionalized(
            RMS_ADD_OP,
            input=allreduce_output,
            residual=residual,
            weight=weight,
            epsilon=self.epsilon)
        quant_out_tuple = auto_functionalized(
            STATIC_FP8_QUANT_OP,
            result=quant_result,
            input=fused_add_rmsnorm_out_tuple[1],
            scale=scale)

        # quant_out, allreduce_output
        return quant_out_tuple[1], fused_add_rmsnorm_out_tuple[2]

    def replacement(quant_result: torch.Tensor, residual: torch.Tensor,
                    input: torch.Tensor, weight: torch.Tensor,
                    scale: torch.Tensor):
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=None,
            quant_out=quant_result,
            scale_out=None,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNormFP8Quant,  # we don't use norm_out afterwards
            scale_factor=scale,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )
        # # quant_out, rms_norm_residual
        return allreduce[4], allreduce[2]

    pm.register_replacement(pattern, replacement, get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusedAddRMSNormStaticQuantNVFP4Pattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (with residual) + static nvfp4 quant with fused flashinfer implementation. Applies to o_proj + rmsnorm after attn + quant and mlp + rmsnorm + quant before attn.

Source code in vllm/compilation/collective_fusion.py 
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
class AllReduceFusedAddRMSNormStaticQuantNVFP4Pattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (with residual)
    + static nvfp4 quant with fused flashinfer implementation.
    Applies to o_proj + rmsnorm after attn + quant and 
    mlp + rmsnorm + quant before attn.
    """

    def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
                 allreduce_params: FlashInferFusedAllReduceParams):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params

    def register(self, pm_pass: PatternMatcherPass):

        def get_inputs():
            input = torch.empty([16, 16], device=self.device, dtype=self.dtype)

            residual = torch.empty([16, 16],
                                   device=self.device,
                                   dtype=self.dtype)
            weight = torch.empty([16, 16],
                                 device=self.device,
                                 dtype=self.dtype)
            quant_result = torch.empty((16, 8),
                                       device=self.device,
                                       dtype=torch.uint8)
            input_global_scale = torch.empty([1, 1],
                                             device=self.device,
                                             dtype=torch.float32)
            output_scale = torch.empty([128, 4],
                                       device=self.device,
                                       dtype=torch.int32)

            return [
                quant_result,
                residual,
                input,
                output_scale,
                weight,
                input_global_scale,
            ]

        def pattern(quant_result: torch.Tensor, residual: torch.Tensor,
                    input: torch.Tensor, output_scale: torch.Tensor,
                    weight: torch.Tensor, input_global_scale: torch.Tensor):
            allreduce_output = tensor_model_parallel_all_reduce(input)

            fused_add_rmsnorm_out_tuple = \
            auto_functionalized(
                RMS_ADD_OP,
                input=allreduce_output,
                residual=residual,
                weight=weight,
                epsilon=self.epsilon)
            quant_out_tuple = auto_functionalized(
                STATIC_FP4_QUANT_OP,
                output=quant_result,
                input=fused_add_rmsnorm_out_tuple[1],
                output_scale=output_scale,
                input_scale=input_global_scale)

            # quant_out, allreduce_output, output_scale
            return quant_out_tuple[1], fused_add_rmsnorm_out_tuple[
                2], quant_out_tuple[2]

        def replacement(quant_result: torch.Tensor, residual: torch.Tensor,
                        input: torch.Tensor, output_scale: torch.Tensor,
                        weight: torch.Tensor,
                        input_global_scale: torch.Tensor):
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=None,
                quant_out=quant_result,
                scale_out=output_scale,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNormFP4Quant,  # we don't use norm_out afterwards
                scale_factor=input_global_scale,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )
            # quant_out, rms_norm_residual, output_scale
            return allreduce[4], allreduce[2], allreduce[5]

        pm.register_replacement(pattern, replacement, get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
985
986
987
988
989
def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
             allreduce_params: FlashInferFusedAllReduceParams):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
def register(self, pm_pass: PatternMatcherPass):

    def get_inputs():
        input = torch.empty([16, 16], device=self.device, dtype=self.dtype)

        residual = torch.empty([16, 16],
                               device=self.device,
                               dtype=self.dtype)
        weight = torch.empty([16, 16],
                             device=self.device,
                             dtype=self.dtype)
        quant_result = torch.empty((16, 8),
                                   device=self.device,
                                   dtype=torch.uint8)
        input_global_scale = torch.empty([1, 1],
                                         device=self.device,
                                         dtype=torch.float32)
        output_scale = torch.empty([128, 4],
                                   device=self.device,
                                   dtype=torch.int32)

        return [
            quant_result,
            residual,
            input,
            output_scale,
            weight,
            input_global_scale,
        ]

    def pattern(quant_result: torch.Tensor, residual: torch.Tensor,
                input: torch.Tensor, output_scale: torch.Tensor,
                weight: torch.Tensor, input_global_scale: torch.Tensor):
        allreduce_output = tensor_model_parallel_all_reduce(input)

        fused_add_rmsnorm_out_tuple = \
        auto_functionalized(
            RMS_ADD_OP,
            input=allreduce_output,
            residual=residual,
            weight=weight,
            epsilon=self.epsilon)
        quant_out_tuple = auto_functionalized(
            STATIC_FP4_QUANT_OP,
            output=quant_result,
            input=fused_add_rmsnorm_out_tuple[1],
            output_scale=output_scale,
            input_scale=input_global_scale)

        # quant_out, allreduce_output, output_scale
        return quant_out_tuple[1], fused_add_rmsnorm_out_tuple[
            2], quant_out_tuple[2]

    def replacement(quant_result: torch.Tensor, residual: torch.Tensor,
                    input: torch.Tensor, output_scale: torch.Tensor,
                    weight: torch.Tensor,
                    input_global_scale: torch.Tensor):
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=None,
            quant_out=quant_result,
            scale_out=output_scale,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNormFP4Quant,  # we don't use norm_out afterwards
            scale_factor=input_global_scale,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )
        # quant_out, rms_norm_residual, output_scale
        return allreduce[4], allreduce[2], allreduce[5]

    pm.register_replacement(pattern, replacement, get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusedRMSNormStaticQuantFP8Pattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (without residual) + static fp8 quant with fused flashinfer implementation. Applies to allreduce + rmsnorm + quant before attn in the first Transformer block.

Source code in vllm/compilation/collective_fusion.py 
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
class AllReduceFusedRMSNormStaticQuantFP8Pattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (without residual) 
    + static fp8 quant with fused flashinfer implementation.
    Applies to allreduce + rmsnorm + quant before attn 
    in the first Transformer block.
    """

    def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
                 allreduce_params: FlashInferFusedAllReduceParams):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params
        self.quant_dtype = torch.float8_e4m3fn

    def register(self, pm_pass: PatternMatcherPass):

        def get_inputs():
            input = torch.zeros([1, 8, 4],
                                device=self.device,
                                dtype=self.dtype)
            rmsnorm_result = torch.empty([1, 8, 4],
                                         device=self.device,
                                         dtype=self.dtype)
            quant_result = torch.empty([1, 8, 4],
                                       device=self.device,
                                       dtype=self.quant_dtype)
            weight = torch.empty([4], device=self.device, dtype=self.dtype)
            scale = torch.tensor(1.0, device=self.device, dtype=torch.float32)
            return [input, rmsnorm_result, quant_result, weight, scale]

        def pattern(
            input: torch.Tensor,
            rmsnorm_result: torch.Tensor,
            quant_result: torch.Tensor,
            weight: torch.Tensor,
            scale: torch.Tensor,
        ):
            all_reduce = tensor_model_parallel_all_reduce(input)
            rmsnorm_out_tuple = auto_functionalized(RMS_OP,
                                                    result=rmsnorm_result,
                                                    input=all_reduce,
                                                    weight=weight,
                                                    epsilon=self.epsilon)

            quant_out_tuple = auto_functionalized(STATIC_FP8_QUANT_OP,
                                                  result=quant_result,
                                                  input=rmsnorm_out_tuple[1],
                                                  scale=scale)

            # quant_out, allreduce_output
            return quant_out_tuple[1], all_reduce

        def replacement(input: torch.Tensor, result_rms: torch.Tensor,
                        quant_result: torch.Tensor, weight: torch.Tensor,
                        scale: torch.Tensor):
            residual = torch.zeros_like(input)
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=result_rms,
                quant_out=quant_result,
                scale_out=None,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNormFP8Quant,  # we don't use norm_out afterwards
                scale_factor=scale,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )

            # quant_out, allreduce_output
            return allreduce[4], allreduce[1]

        pm.register_replacement(pattern, replacement, get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

quant_dtype instance-attribute 

quant_dtype = float8_e4m3fn

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
726
727
728
729
730
731
def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
             allreduce_params: FlashInferFusedAllReduceParams):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params
    self.quant_dtype = torch.float8_e4m3fn

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
def register(self, pm_pass: PatternMatcherPass):

    def get_inputs():
        input = torch.zeros([1, 8, 4],
                            device=self.device,
                            dtype=self.dtype)
        rmsnorm_result = torch.empty([1, 8, 4],
                                     device=self.device,
                                     dtype=self.dtype)
        quant_result = torch.empty([1, 8, 4],
                                   device=self.device,
                                   dtype=self.quant_dtype)
        weight = torch.empty([4], device=self.device, dtype=self.dtype)
        scale = torch.tensor(1.0, device=self.device, dtype=torch.float32)
        return [input, rmsnorm_result, quant_result, weight, scale]

    def pattern(
        input: torch.Tensor,
        rmsnorm_result: torch.Tensor,
        quant_result: torch.Tensor,
        weight: torch.Tensor,
        scale: torch.Tensor,
    ):
        all_reduce = tensor_model_parallel_all_reduce(input)
        rmsnorm_out_tuple = auto_functionalized(RMS_OP,
                                                result=rmsnorm_result,
                                                input=all_reduce,
                                                weight=weight,
                                                epsilon=self.epsilon)

        quant_out_tuple = auto_functionalized(STATIC_FP8_QUANT_OP,
                                              result=quant_result,
                                              input=rmsnorm_out_tuple[1],
                                              scale=scale)

        # quant_out, allreduce_output
        return quant_out_tuple[1], all_reduce

    def replacement(input: torch.Tensor, result_rms: torch.Tensor,
                    quant_result: torch.Tensor, weight: torch.Tensor,
                    scale: torch.Tensor):
        residual = torch.zeros_like(input)
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=result_rms,
            quant_out=quant_result,
            scale_out=None,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNormFP8Quant,  # we don't use norm_out afterwards
            scale_factor=scale,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )

        # quant_out, allreduce_output
        return allreduce[4], allreduce[1]

    pm.register_replacement(pattern, replacement, get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusedRMSNormStaticQuantNVFP4Pattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (without residual) + static nvfp4 quant with fused flashinfer implementation. Applies to allreduce + rmsnorm + quant before attn in the first Transformer block.

Source code in vllm/compilation/collective_fusion.py 
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
class AllReduceFusedRMSNormStaticQuantNVFP4Pattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (without residual) 
    + static nvfp4 quant with fused flashinfer implementation.
    Applies to allreduce + rmsnorm + quant before attn 
    in the first Transformer block.
    """

    def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
                 allreduce_params: FlashInferFusedAllReduceParams):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params

    def register(self, pm_pass: PatternMatcherPass):

        def get_inputs():
            input = torch.empty([1, 16, 16],
                                device=self.device,
                                dtype=self.dtype)

            rmsnorm_result = torch.empty([1, 16, 16],
                                         device=self.device,
                                         dtype=self.dtype)
            quant_result = torch.empty((16, 8),
                                       device=self.device,
                                       dtype=torch.uint8)
            input_global_scale = torch.empty([1, 1],
                                             device=self.device,
                                             dtype=torch.float32)
            weight = torch.empty([16], device=self.device, dtype=self.dtype)
            output_scale = torch.empty([128, 4],
                                       device=self.device,
                                       dtype=torch.int32)

            return [
                input, rmsnorm_result, quant_result, weight,
                input_global_scale, output_scale
            ]

        def pattern(
            input: torch.Tensor,
            rmsnorm_result: torch.Tensor,
            quant_result: torch.Tensor,
            weight: torch.Tensor,
            input_global_scale: torch.Tensor,
            output_scale: torch.Tensor,
        ):
            all_reduce = tensor_model_parallel_all_reduce(input)
            rmsnorm_out_tuple = auto_functionalized(RMS_OP,
                                                    result=rmsnorm_result,
                                                    input=all_reduce,
                                                    weight=weight,
                                                    epsilon=self.epsilon)

            quant_out_tuple = auto_functionalized(
                STATIC_FP4_QUANT_OP,
                output=quant_result,
                input=rmsnorm_out_tuple[1],
                output_scale=output_scale,
                input_scale=input_global_scale)

            # quant_out, allreduce_output, output_scale
            return quant_out_tuple[1], all_reduce, quant_out_tuple[2]

        def replacement(input: torch.Tensor, result_rms: torch.Tensor,
                        quant_result: torch.Tensor, weight: torch.Tensor,
                        input_global_scale: torch.Tensor,
                        output_scale: torch.Tensor):
            residual = torch.zeros_like(input)
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=result_rms,
                quant_out=quant_result,
                scale_out=output_scale,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNormFP4Quant,  # we don't use norm_out afterwards
                scale_factor=input_global_scale,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )

            # quant_out, allreduce_output, output_scale
            return allreduce[4], allreduce[1], allreduce[5]

        pm.register_replacement(pattern, replacement, get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
893
894
895
896
897
def __init__(self, epsilon: float, dtype: torch.dtype, device: str,
             allreduce_params: FlashInferFusedAllReduceParams):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
def register(self, pm_pass: PatternMatcherPass):

    def get_inputs():
        input = torch.empty([1, 16, 16],
                            device=self.device,
                            dtype=self.dtype)

        rmsnorm_result = torch.empty([1, 16, 16],
                                     device=self.device,
                                     dtype=self.dtype)
        quant_result = torch.empty((16, 8),
                                   device=self.device,
                                   dtype=torch.uint8)
        input_global_scale = torch.empty([1, 1],
                                         device=self.device,
                                         dtype=torch.float32)
        weight = torch.empty([16], device=self.device, dtype=self.dtype)
        output_scale = torch.empty([128, 4],
                                   device=self.device,
                                   dtype=torch.int32)

        return [
            input, rmsnorm_result, quant_result, weight,
            input_global_scale, output_scale
        ]

    def pattern(
        input: torch.Tensor,
        rmsnorm_result: torch.Tensor,
        quant_result: torch.Tensor,
        weight: torch.Tensor,
        input_global_scale: torch.Tensor,
        output_scale: torch.Tensor,
    ):
        all_reduce = tensor_model_parallel_all_reduce(input)
        rmsnorm_out_tuple = auto_functionalized(RMS_OP,
                                                result=rmsnorm_result,
                                                input=all_reduce,
                                                weight=weight,
                                                epsilon=self.epsilon)

        quant_out_tuple = auto_functionalized(
            STATIC_FP4_QUANT_OP,
            output=quant_result,
            input=rmsnorm_out_tuple[1],
            output_scale=output_scale,
            input_scale=input_global_scale)

        # quant_out, allreduce_output, output_scale
        return quant_out_tuple[1], all_reduce, quant_out_tuple[2]

    def replacement(input: torch.Tensor, result_rms: torch.Tensor,
                    quant_result: torch.Tensor, weight: torch.Tensor,
                    input_global_scale: torch.Tensor,
                    output_scale: torch.Tensor):
        residual = torch.zeros_like(input)
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=result_rms,
            quant_out=quant_result,
            scale_out=output_scale,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNormFP4Quant,  # we don't use norm_out afterwards
            scale_factor=input_global_scale,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )

        # quant_out, allreduce_output, output_scale
        return allreduce[4], allreduce[1], allreduce[5]

    pm.register_replacement(pattern, replacement, get_inputs(),
                            pm.fwd_only, pm_pass)

AllReduceFusionPass

Bases: VllmInductorPass

Source code in vllm/compilation/collective_fusion.py 
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
class AllReduceFusionPass(VllmInductorPass):

    def __init__(self, config: VllmConfig):
        super().__init__(config)
        self.disabled = True
        self.tp_size = get_tensor_model_parallel_world_size()
        if self.tp_size <= 1:
            return
        self.patterns: PatternMatcherPass = PatternMatcherPass(
            pass_name="all_reduce_fusion_pass")
        if config.model_config is None:
            return
        self.hidden_dim = config.model_config.get_hidden_size()
        self.group = get_tp_group().device_group
        rank = get_tensor_model_parallel_rank()
        use_fp32_lamport = self.model_dtype == torch.float32
        if flashinfer_comm is None:
            logger.warning(
                "Flashinfer is not installed or comm module not found, "
                "skipping allreduce fusion pass")
            return
        # Check if the world size is supported
        if self.tp_size not in _FI_MAX_SIZES:
            logger.warning(
                "Flashinfer allreduce fusion is not "
                "supported for world size %s",
                self.tp_size,
            )
            return
        max_num_token = min(
            _FI_MAX_SIZES.get(self.tp_size, _DEFAULT_FI_MAX_SIZE) //
            (self.hidden_dim * self.tp_size * (4 if use_fp32_lamport else 2)),
            config.compilation_config.pass_config.
            fi_allreduce_fusion_max_token_num)
        self.ipc_handles, workspace_tensor = (
            flashinfer_comm.trtllm_create_ipc_workspace_for_all_reduce_fusion(
                tp_rank=rank,
                tp_size=self.tp_size,
                max_token_num=max_num_token,
                hidden_dim=self.hidden_dim,
                group=self.group,
                use_fp32_lamport=use_fp32_lamport,
            ))

        global _FI_WORKSPACE_TENSOR
        _FI_WORKSPACE_TENSOR = workspace_tensor
        self.allreduce_params = FlashInferFusedAllReduceParams(
            rank=rank,
            world_size=self.tp_size,
            use_fp32_lamport=use_fp32_lamport,
            max_token_num=max_num_token,
            # fuse rms norm static fp8 quant fused op
            # in fallback path, when we don't use flashinfer
            fuse_rms_quant=config.compilation_config.pass_config.enable_fusion)

        for epsilon in [1e-5, 1e-6]:
            AllReduceFusedRMSNormStaticQuantFP8Pattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)
            AllReduceFusedAddRMSNormStaticQuantFP8Pattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)
            if current_platform.has_device_capability(100):
                AllReduceFusedRMSNormStaticQuantNVFP4Pattern(
                    epsilon,
                    self.model_dtype,
                    self.device,
                    self.allreduce_params,
                ).register(self.patterns)
                AllReduceFusedAddRMSNormStaticQuantNVFP4Pattern(
                    epsilon,
                    self.model_dtype,
                    self.device,
                    self.allreduce_params,
                ).register(self.patterns)
            AllReduceRMSNormPattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)
            AllReduceFusedAddRMSNormPattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)

            # WARNING: This is a hack to clear the pattern matcher cache
            # and allow multiple values of epsilon.
            torch._inductor.pattern_matcher._seen_patterns.clear()

        self.disabled = False

    def __call__(self, graph: fx.Graph):
        if self.disabled:
            return
        self.begin()
        self.dump_graph(graph, "before_all_reduce_fusion_pass")
        count = self.patterns.apply(graph)
        logger.debug("Replaced %s patterns", count)
        self.dump_graph(graph, "after_all_reduce_fusion_pass")
        self.end_and_log()

    def __del__(self):
        if self.disabled:
            return
        if flashinfer_comm is not None:
            flashinfer_comm.trtllm_destroy_ipc_workspace_for_all_reduce(
                self.ipc_handles, self.group)

allreduce_params instance-attribute 

allreduce_params = FlashInferFusedAllReduceParams(
    rank=rank,
    world_size=tp_size,
    use_fp32_lamport=use_fp32_lamport,
    max_token_num=max_num_token,
    fuse_rms_quant=enable_fusion,
)

disabled instance-attribute 

disabled = False

group instance-attribute 

group = device_group

hidden_dim instance-attribute 

hidden_dim = get_hidden_size()

patterns instance-attribute 

patterns: PatternMatcherPass = PatternMatcherPass(
    pass_name="all_reduce_fusion_pass"
)

tp_size instance-attribute 

tp_size = get_tensor_model_parallel_world_size()

__call__ 

__call__(graph: Graph)
Source code in vllm/compilation/collective_fusion.py 
1169
1170
1171
1172
1173
1174
1175
1176
1177
def __call__(self, graph: fx.Graph):
    if self.disabled:
        return
    self.begin()
    self.dump_graph(graph, "before_all_reduce_fusion_pass")
    count = self.patterns.apply(graph)
    logger.debug("Replaced %s patterns", count)
    self.dump_graph(graph, "after_all_reduce_fusion_pass")
    self.end_and_log()

__del__ 

__del__()
Source code in vllm/compilation/collective_fusion.py 
1179
1180
1181
1182
1183
1184
def __del__(self):
    if self.disabled:
        return
    if flashinfer_comm is not None:
        flashinfer_comm.trtllm_destroy_ipc_workspace_for_all_reduce(
            self.ipc_handles, self.group)

__init__ 

__init__(config: VllmConfig)
Source code in vllm/compilation/collective_fusion.py 
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
def __init__(self, config: VllmConfig):
    super().__init__(config)
    self.disabled = True
    self.tp_size = get_tensor_model_parallel_world_size()
    if self.tp_size <= 1:
        return
    self.patterns: PatternMatcherPass = PatternMatcherPass(
        pass_name="all_reduce_fusion_pass")
    if config.model_config is None:
        return
    self.hidden_dim = config.model_config.get_hidden_size()
    self.group = get_tp_group().device_group
    rank = get_tensor_model_parallel_rank()
    use_fp32_lamport = self.model_dtype == torch.float32
    if flashinfer_comm is None:
        logger.warning(
            "Flashinfer is not installed or comm module not found, "
            "skipping allreduce fusion pass")
        return
    # Check if the world size is supported
    if self.tp_size not in _FI_MAX_SIZES:
        logger.warning(
            "Flashinfer allreduce fusion is not "
            "supported for world size %s",
            self.tp_size,
        )
        return
    max_num_token = min(
        _FI_MAX_SIZES.get(self.tp_size, _DEFAULT_FI_MAX_SIZE) //
        (self.hidden_dim * self.tp_size * (4 if use_fp32_lamport else 2)),
        config.compilation_config.pass_config.
        fi_allreduce_fusion_max_token_num)
    self.ipc_handles, workspace_tensor = (
        flashinfer_comm.trtllm_create_ipc_workspace_for_all_reduce_fusion(
            tp_rank=rank,
            tp_size=self.tp_size,
            max_token_num=max_num_token,
            hidden_dim=self.hidden_dim,
            group=self.group,
            use_fp32_lamport=use_fp32_lamport,
        ))

    global _FI_WORKSPACE_TENSOR
    _FI_WORKSPACE_TENSOR = workspace_tensor
    self.allreduce_params = FlashInferFusedAllReduceParams(
        rank=rank,
        world_size=self.tp_size,
        use_fp32_lamport=use_fp32_lamport,
        max_token_num=max_num_token,
        # fuse rms norm static fp8 quant fused op
        # in fallback path, when we don't use flashinfer
        fuse_rms_quant=config.compilation_config.pass_config.enable_fusion)

    for epsilon in [1e-5, 1e-6]:
        AllReduceFusedRMSNormStaticQuantFP8Pattern(
            epsilon,
            self.model_dtype,
            self.device,
            self.allreduce_params,
        ).register(self.patterns)
        AllReduceFusedAddRMSNormStaticQuantFP8Pattern(
            epsilon,
            self.model_dtype,
            self.device,
            self.allreduce_params,
        ).register(self.patterns)
        if current_platform.has_device_capability(100):
            AllReduceFusedRMSNormStaticQuantNVFP4Pattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)
            AllReduceFusedAddRMSNormStaticQuantNVFP4Pattern(
                epsilon,
                self.model_dtype,
                self.device,
                self.allreduce_params,
            ).register(self.patterns)
        AllReduceRMSNormPattern(
            epsilon,
            self.model_dtype,
            self.device,
            self.allreduce_params,
        ).register(self.patterns)
        AllReduceFusedAddRMSNormPattern(
            epsilon,
            self.model_dtype,
            self.device,
            self.allreduce_params,
        ).register(self.patterns)

        # WARNING: This is a hack to clear the pattern matcher cache
        # and allow multiple values of epsilon.
        torch._inductor.pattern_matcher._seen_patterns.clear()

    self.disabled = False

AllReduceRMSNormPattern

Bases: BasePattern

This pattern replaces the allreduce + rms norm (without residual) with fused flashinfer implementation. Applies to allreduce + rmsnorm before attn in the first Transformer block.

Source code in vllm/compilation/collective_fusion.py 
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
class AllReduceRMSNormPattern(BasePattern):
    """
    This pattern replaces the allreduce + rms norm (without residual) 
    with fused flashinfer implementation.
    Applies to allreduce + rmsnorm before attn in the first Transformer block.
    """

    def __init__(
        self,
        epsilon: float,
        dtype: torch.dtype,
        device: str,
        allreduce_params: FlashInferFusedAllReduceParams,
    ):
        super().__init__(dtype, device)
        self.epsilon = epsilon
        self.allreduce_params = allreduce_params

    def get_inputs(self):
        input = torch.empty([1, 8, 4], device=self.device, dtype=self.dtype)
        rms_result = torch.empty([1, 8, 4],
                                 device=self.device,
                                 dtype=self.dtype)
        weight = torch.empty([4], device=self.device, dtype=self.dtype)

        return [input, rms_result, weight]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(input: torch.Tensor, rms_result: torch.Tensor,
                    weight: torch.Tensor):
            allreduce_output = tensor_model_parallel_all_reduce(input)
            rms = auto_functionalized(
                RMS_OP,
                result=rms_result,
                input=allreduce_output,
                weight=weight,
                epsilon=self.epsilon,
            )
            # rms_result, allreduce_output
            return rms[1], allreduce_output

        def replacement(input: torch.Tensor, rms_result: torch.Tensor,
                        weight: torch.Tensor):
            residual = torch.zeros_like(input)
            allreduce = auto_functionalized(
                flashinfer_trtllm_fused_allreduce_norm,
                allreduce_in=input,
                residual=residual,
                norm_out=rms_result,
                quant_out=None,
                scale_out=None,
                rms_gamma=weight,
                rms_eps=self.epsilon,
                pattern_code=flashinfer_comm.AllReduceFusionPattern.
                kARResidualRMSNorm,
                **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
            )
            # rms_result, allreduce_in
            return allreduce[3], allreduce[1]

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

allreduce_params instance-attribute 

allreduce_params = allreduce_params

epsilon instance-attribute 

epsilon = epsilon

__init__ 

__init__(
    epsilon: float,
    dtype: dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
)
Source code in vllm/compilation/collective_fusion.py 
595
596
597
598
599
600
601
602
603
604
def __init__(
    self,
    epsilon: float,
    dtype: torch.dtype,
    device: str,
    allreduce_params: FlashInferFusedAllReduceParams,
):
    super().__init__(dtype, device)
    self.epsilon = epsilon
    self.allreduce_params = allreduce_params

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
606
607
608
609
610
611
612
613
def get_inputs(self):
    input = torch.empty([1, 8, 4], device=self.device, dtype=self.dtype)
    rms_result = torch.empty([1, 8, 4],
                             device=self.device,
                             dtype=self.dtype)
    weight = torch.empty([4], device=self.device, dtype=self.dtype)

    return [input, rms_result, weight]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
def register(self, pm_pass: PatternMatcherPass):

    def pattern(input: torch.Tensor, rms_result: torch.Tensor,
                weight: torch.Tensor):
        allreduce_output = tensor_model_parallel_all_reduce(input)
        rms = auto_functionalized(
            RMS_OP,
            result=rms_result,
            input=allreduce_output,
            weight=weight,
            epsilon=self.epsilon,
        )
        # rms_result, allreduce_output
        return rms[1], allreduce_output

    def replacement(input: torch.Tensor, rms_result: torch.Tensor,
                    weight: torch.Tensor):
        residual = torch.zeros_like(input)
        allreduce = auto_functionalized(
            flashinfer_trtllm_fused_allreduce_norm,
            allreduce_in=input,
            residual=residual,
            norm_out=rms_result,
            quant_out=None,
            scale_out=None,
            rms_gamma=weight,
            rms_eps=self.epsilon,
            pattern_code=flashinfer_comm.AllReduceFusionPattern.
            kARResidualRMSNorm,
            **self.allreduce_params.get_trtllm_fused_allreduce_kwargs(),
        )
        # rms_result, allreduce_in
        return allreduce[3], allreduce[1]

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

AsyncTPPass

Bases: VllmInductorPass

Source code in vllm/compilation/collective_fusion.py 
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
class AsyncTPPass(VllmInductorPass):

    def __init__(self, config: VllmConfig):
        super().__init__(config)

        # Enable symmetric memory for the TP process group
        enable_symm_mem_for_group(get_tp_group().device_group.group_name)
        self.patterns: PatternMatcherPass = PatternMatcherPass(
            pass_name="async_tp_pass")
        GEMMReduceScatterPattern(self.model_dtype,
                                 self.device).register(self.patterns)

        AllGatherGEMMPattern(self.model_dtype,
                             self.device).register(self.patterns)

        # These fusions are enabled only for bfloat16 models because
        # `scaled_mm` or `cutlass_scaled_mm` with per-token (row-wise) scaling
        # only supports bfloat16 as the output dtype.
        if self.model_dtype == torch.bfloat16:
            ScaledMMReduceScatterPattern(self.model_dtype,
                                         self.device).register(self.patterns)
            AllGatherScaledMMPattern(self.model_dtype,
                                     self.device).register(self.patterns)

            CutlassScaledMMReduceScatterPattern(
                self.model_dtype, self.device).register(self.patterns)
            AllGatherCutlassScaledMMPattern(
                self.model_dtype, self.device).register(self.patterns)

    def is_applicable_for_shape(self, shape: Optional[int]) -> bool:
        # only do replace for specific shapes
        tp_size = get_tensor_model_parallel_world_size()
        return shape is not None and shape % tp_size == 0

    def __call__(self, graph: fx.Graph):
        self.begin()
        self.dump_graph(graph, "before_async_tp_pass")
        count = self.patterns.apply(graph)
        logger.debug("Replaced %s patterns with async TP pass.", count)
        self.dump_graph(graph, "after_async_tp_pass")
        self.end_and_log()

patterns instance-attribute 

patterns: PatternMatcherPass = PatternMatcherPass(
    pass_name="async_tp_pass"
)

__call__ 

__call__(graph: Graph)
Source code in vllm/compilation/collective_fusion.py 
384
385
386
387
388
389
390
def __call__(self, graph: fx.Graph):
    self.begin()
    self.dump_graph(graph, "before_async_tp_pass")
    count = self.patterns.apply(graph)
    logger.debug("Replaced %s patterns with async TP pass.", count)
    self.dump_graph(graph, "after_async_tp_pass")
    self.end_and_log()

__init__ 

__init__(config: VllmConfig)
Source code in vllm/compilation/collective_fusion.py 
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
def __init__(self, config: VllmConfig):
    super().__init__(config)

    # Enable symmetric memory for the TP process group
    enable_symm_mem_for_group(get_tp_group().device_group.group_name)
    self.patterns: PatternMatcherPass = PatternMatcherPass(
        pass_name="async_tp_pass")
    GEMMReduceScatterPattern(self.model_dtype,
                             self.device).register(self.patterns)

    AllGatherGEMMPattern(self.model_dtype,
                         self.device).register(self.patterns)

    # These fusions are enabled only for bfloat16 models because
    # `scaled_mm` or `cutlass_scaled_mm` with per-token (row-wise) scaling
    # only supports bfloat16 as the output dtype.
    if self.model_dtype == torch.bfloat16:
        ScaledMMReduceScatterPattern(self.model_dtype,
                                     self.device).register(self.patterns)
        AllGatherScaledMMPattern(self.model_dtype,
                                 self.device).register(self.patterns)

        CutlassScaledMMReduceScatterPattern(
            self.model_dtype, self.device).register(self.patterns)
        AllGatherCutlassScaledMMPattern(
            self.model_dtype, self.device).register(self.patterns)

is_applicable_for_shape 

is_applicable_for_shape(shape: Optional[int]) -> bool
Source code in vllm/compilation/collective_fusion.py 
379
380
381
382
def is_applicable_for_shape(self, shape: Optional[int]) -> bool:
    # only do replace for specific shapes
    tp_size = get_tensor_model_parallel_world_size()
    return shape is not None and shape % tp_size == 0

BasePattern

Source code in vllm/compilation/collective_fusion.py 
45
46
47
48
49
50
51
class BasePattern:

    def __init__(self, dtype: torch.dtype, device: str):
        self.dtype = dtype
        self.device = device
        self.tp = get_tp_group()
        self.tp_size = get_tensor_model_parallel_world_size()

device instance-attribute 

device = device

dtype instance-attribute 

dtype = dtype

tp instance-attribute 

tp = get_tp_group()

tp_size instance-attribute 

tp_size = get_tensor_model_parallel_world_size()

__init__ 

__init__(dtype: dtype, device: str)
Source code in vllm/compilation/collective_fusion.py 
47
48
49
50
51
def __init__(self, dtype: torch.dtype, device: str):
    self.dtype = dtype
    self.device = device
    self.tp = get_tp_group()
    self.tp_size = get_tensor_model_parallel_world_size()

CutlassScaledMMReduceScatterPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
class CutlassScaledMMReduceScatterPattern(BasePattern):

    def get_inputs(self):
        input = torch.empty([16, 16], device=self.device, dtype=FP8_DTYPE)
        mm_weight = torch.empty([16, 16], device=self.device,
                                dtype=FP8_DTYPE).contiguous().transpose(0, 1)
        scale_a = torch.empty([16, 1], device=self.device, dtype=torch.float32)
        scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

        cutlass_mm_output = torch.empty([16, 16],
                                        device=self.device,
                                        dtype=self.dtype)
        return [input, mm_weight, scale_a, scale_b, cutlass_mm_output]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(input: torch.Tensor, weight: torch.Tensor,
                    scale_a: torch.Tensor, scale_b: torch.Tensor,
                    cutlass_mm_output: torch.Tensor) -> torch.Tensor:
            cutlass_scaled_mm = torch.ops.higher_order.auto_functionalized(
                torch.ops._C.cutlass_scaled_mm.default,
                out=cutlass_mm_output,
                a=input,
                b=weight,
                a_scales=scale_a,
                b_scales=scale_b,
                bias=None)

            reduce_scatter = torch.ops.vllm.reduce_scatter.default(
                cutlass_scaled_mm[1],
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name)
            return reduce_scatter

        def replacement(input: torch.Tensor, mat2: torch.Tensor,
                        scale_a: torch.Tensor, scale_b: torch.Tensor,
                        cutlass_mm_output: torch.Tensor) -> torch.Tensor:
            gemm_rs = torch.ops.symm_mem.fused_scaled_matmul_reduce_scatter(
                input,
                mat2,
                scale_a,
                scale_b,
                "avg",
                scatter_dim=0,
                out_dtype=self.dtype,
                group_name=self.tp.device_group.group_name,
            )

            return gemm_rs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
234
235
236
237
238
239
240
241
242
243
244
def get_inputs(self):
    input = torch.empty([16, 16], device=self.device, dtype=FP8_DTYPE)
    mm_weight = torch.empty([16, 16], device=self.device,
                            dtype=FP8_DTYPE).contiguous().transpose(0, 1)
    scale_a = torch.empty([16, 1], device=self.device, dtype=torch.float32)
    scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)

    cutlass_mm_output = torch.empty([16, 16],
                                    device=self.device,
                                    dtype=self.dtype)
    return [input, mm_weight, scale_a, scale_b, cutlass_mm_output]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
def register(self, pm_pass: PatternMatcherPass):

    def pattern(input: torch.Tensor, weight: torch.Tensor,
                scale_a: torch.Tensor, scale_b: torch.Tensor,
                cutlass_mm_output: torch.Tensor) -> torch.Tensor:
        cutlass_scaled_mm = torch.ops.higher_order.auto_functionalized(
            torch.ops._C.cutlass_scaled_mm.default,
            out=cutlass_mm_output,
            a=input,
            b=weight,
            a_scales=scale_a,
            b_scales=scale_b,
            bias=None)

        reduce_scatter = torch.ops.vllm.reduce_scatter.default(
            cutlass_scaled_mm[1],
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name)
        return reduce_scatter

    def replacement(input: torch.Tensor, mat2: torch.Tensor,
                    scale_a: torch.Tensor, scale_b: torch.Tensor,
                    cutlass_mm_output: torch.Tensor) -> torch.Tensor:
        gemm_rs = torch.ops.symm_mem.fused_scaled_matmul_reduce_scatter(
            input,
            mat2,
            scale_a,
            scale_b,
            "avg",
            scatter_dim=0,
            out_dtype=self.dtype,
            group_name=self.tp.device_group.group_name,
        )

        return gemm_rs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

FlashInferFusedAllReduceParams

Parameters for FlashInfer fused allreduce operations.

Source code in vllm/compilation/collective_fusion.py 
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
class FlashInferFusedAllReduceParams:
    """Parameters for FlashInfer fused allreduce operations."""

    def __init__(
        self,
        rank: int,
        world_size: int,
        use_fp32_lamport: bool = False,
        max_token_num: int = 1024,
        fuse_rms_quant: bool = False,
    ):
        self.rank = rank
        self.world_size = world_size
        self.use_fp32_lamport = use_fp32_lamport
        self.trigger_completion_at_end = True
        self.launch_with_pdl = True
        self.fp32_acc = True
        self.use_oneshot = False
        self.max_token_num = max_token_num
        self.fuse_rms_quant = fuse_rms_quant

    def get_trtllm_fused_allreduce_kwargs(self):
        return {
            "world_rank": self.rank,
            "world_size": self.world_size,
            "launch_with_pdl": self.launch_with_pdl,
            "trigger_completion_at_end": self.trigger_completion_at_end,
            "fp32_acc": self.fp32_acc,
            "max_token_num": self.max_token_num,
            "fuse_rms_quant": self.fuse_rms_quant,
        }

fp32_acc instance-attribute 

fp32_acc = True

fuse_rms_quant instance-attribute 

fuse_rms_quant = fuse_rms_quant

launch_with_pdl instance-attribute 

launch_with_pdl = True

max_token_num instance-attribute 

max_token_num = max_token_num

rank instance-attribute 

rank = rank

trigger_completion_at_end instance-attribute 

trigger_completion_at_end = True

use_fp32_lamport instance-attribute 

use_fp32_lamport = use_fp32_lamport

use_oneshot instance-attribute 

use_oneshot = False

world_size instance-attribute 

world_size = world_size

__init__ 

__init__(
    rank: int,
    world_size: int,
    use_fp32_lamport: bool = False,
    max_token_num: int = 1024,
    fuse_rms_quant: bool = False,
)
Source code in vllm/compilation/collective_fusion.py 
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
def __init__(
    self,
    rank: int,
    world_size: int,
    use_fp32_lamport: bool = False,
    max_token_num: int = 1024,
    fuse_rms_quant: bool = False,
):
    self.rank = rank
    self.world_size = world_size
    self.use_fp32_lamport = use_fp32_lamport
    self.trigger_completion_at_end = True
    self.launch_with_pdl = True
    self.fp32_acc = True
    self.use_oneshot = False
    self.max_token_num = max_token_num
    self.fuse_rms_quant = fuse_rms_quant

get_trtllm_fused_allreduce_kwargs 

get_trtllm_fused_allreduce_kwargs()
Source code in vllm/compilation/collective_fusion.py 
576
577
578
579
580
581
582
583
584
585
def get_trtllm_fused_allreduce_kwargs(self):
    return {
        "world_rank": self.rank,
        "world_size": self.world_size,
        "launch_with_pdl": self.launch_with_pdl,
        "trigger_completion_at_end": self.trigger_completion_at_end,
        "fp32_acc": self.fp32_acc,
        "max_token_num": self.max_token_num,
        "fuse_rms_quant": self.fuse_rms_quant,
    }

GEMMReduceScatterPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
class GEMMReduceScatterPattern(BasePattern):

    def get_inputs(self):
        mul = torch.empty([16, 4], device=self.device, dtype=self.dtype)
        mm_weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
        return [mul, mm_weight]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(mul: torch.Tensor, mm_weight: torch.Tensor):
            mm = torch.ops.aten.mm.default(mul, mm_weight)
            reduce_scatter = torch.ops.vllm.reduce_scatter.default(
                mm,
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name,
            )
            return reduce_scatter

        def replacement(mul: torch.Tensor, mm_weight: torch.Tensor):
            gemm_rs = torch.ops.symm_mem.fused_matmul_reduce_scatter(
                mul,
                mm_weight,
                "avg",
                scatter_dim=0,
                group_name=self.tp.device_group.group_name,
            )

            return gemm_rs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
56
57
58
59
def get_inputs(self):
    mul = torch.empty([16, 4], device=self.device, dtype=self.dtype)
    mm_weight = torch.empty([4, 4], device=self.device, dtype=self.dtype)
    return [mul, mm_weight]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
def register(self, pm_pass: PatternMatcherPass):

    def pattern(mul: torch.Tensor, mm_weight: torch.Tensor):
        mm = torch.ops.aten.mm.default(mul, mm_weight)
        reduce_scatter = torch.ops.vllm.reduce_scatter.default(
            mm,
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name,
        )
        return reduce_scatter

    def replacement(mul: torch.Tensor, mm_weight: torch.Tensor):
        gemm_rs = torch.ops.symm_mem.fused_matmul_reduce_scatter(
            mul,
            mm_weight,
            "avg",
            scatter_dim=0,
            group_name=self.tp.device_group.group_name,
        )

        return gemm_rs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

ScaledMMReduceScatterPattern

Bases: BasePattern

Source code in vllm/compilation/collective_fusion.py 
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
class ScaledMMReduceScatterPattern(BasePattern):

    def get_inputs(self):
        input = torch.empty([16, 16], device=self.device, dtype=FP8_DTYPE)
        mm_weight = torch.empty([16, 16], device=self.device,
                                dtype=FP8_DTYPE).contiguous().transpose(0, 1)
        scale_a = torch.empty([16, 1], device=self.device, dtype=torch.float32)
        scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)
        return [input, mm_weight, scale_a, scale_b]

    def register(self, pm_pass: PatternMatcherPass):

        def pattern(input: torch.Tensor, mat2: torch.Tensor,
                    scale_a: torch.Tensor,
                    scale_b: torch.Tensor) -> torch.Tensor:
            scaled_mm = torch.ops.aten._scaled_mm.default(input,
                                                          mat2=mat2,
                                                          scale_a=scale_a,
                                                          scale_b=scale_b,
                                                          bias=None,
                                                          scale_result=None,
                                                          out_dtype=self.dtype)
            reduce_scatter = torch.ops.vllm.reduce_scatter.default(
                scaled_mm,
                dim=0,
                world_size=self.tp_size,
                group_name=self.tp.unique_name)
            return reduce_scatter

        def replacement(input: torch.Tensor, mat2: torch.Tensor,
                        scale_a: torch.Tensor,
                        scale_b: torch.Tensor) -> torch.Tensor:
            gemm_rs = torch.ops.symm_mem.fused_scaled_matmul_reduce_scatter(
                input,
                mat2,
                scale_a,
                scale_b,
                "avg",
                scatter_dim=0,
                out_dtype=self.dtype,
                group_name=self.tp.device_group.group_name,
            )

            return gemm_rs

        pm.register_replacement(pattern, replacement, self.get_inputs(),
                                pm.fwd_only, pm_pass)

get_inputs 

get_inputs()
Source code in vllm/compilation/collective_fusion.py 
128
129
130
131
132
133
134
def get_inputs(self):
    input = torch.empty([16, 16], device=self.device, dtype=FP8_DTYPE)
    mm_weight = torch.empty([16, 16], device=self.device,
                            dtype=FP8_DTYPE).contiguous().transpose(0, 1)
    scale_a = torch.empty([16, 1], device=self.device, dtype=torch.float32)
    scale_b = torch.empty([1, 16], device=self.device, dtype=torch.float32)
    return [input, mm_weight, scale_a, scale_b]

register 

register(pm_pass: PatternMatcherPass)
Source code in vllm/compilation/collective_fusion.py 
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
def register(self, pm_pass: PatternMatcherPass):

    def pattern(input: torch.Tensor, mat2: torch.Tensor,
                scale_a: torch.Tensor,
                scale_b: torch.Tensor) -> torch.Tensor:
        scaled_mm = torch.ops.aten._scaled_mm.default(input,
                                                      mat2=mat2,
                                                      scale_a=scale_a,
                                                      scale_b=scale_b,
                                                      bias=None,
                                                      scale_result=None,
                                                      out_dtype=self.dtype)
        reduce_scatter = torch.ops.vllm.reduce_scatter.default(
            scaled_mm,
            dim=0,
            world_size=self.tp_size,
            group_name=self.tp.unique_name)
        return reduce_scatter

    def replacement(input: torch.Tensor, mat2: torch.Tensor,
                    scale_a: torch.Tensor,
                    scale_b: torch.Tensor) -> torch.Tensor:
        gemm_rs = torch.ops.symm_mem.fused_scaled_matmul_reduce_scatter(
            input,
            mat2,
            scale_a,
            scale_b,
            "avg",
            scatter_dim=0,
            out_dtype=self.dtype,
            group_name=self.tp.device_group.group_name,
        )

        return gemm_rs

    pm.register_replacement(pattern, replacement, self.get_inputs(),
                            pm.fwd_only, pm_pass)

call_trtllm_fused_allreduce_norm 

call_trtllm_fused_allreduce_norm(
    allreduce_in: Tensor,
    residual: Tensor,
    rms_gamma: Tensor,
    rms_eps: float,
    world_rank: int,
    world_size: int,
    launch_with_pdl: bool,
    trigger_completion_at_end: bool,
    fp32_acc: bool,
    max_token_num: int,
    pattern_code: int,
    fuse_rms_quant: bool,
    norm_out: Optional[Tensor] = None,
    quant_out: Optional[Tensor] = None,
    scale_out: Optional[Tensor] = None,
    scale_factor: Optional[Tensor] = None,
) -> None
Source code in vllm/compilation/collective_fusion.py 
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
def call_trtllm_fused_allreduce_norm(
    allreduce_in: torch.Tensor,
    residual: torch.Tensor,
    rms_gamma: torch.Tensor,
    rms_eps: float,
    world_rank: int,
    world_size: int,
    launch_with_pdl: bool,
    trigger_completion_at_end: bool,
    fp32_acc: bool,
    max_token_num: int,
    pattern_code: int,
    fuse_rms_quant: bool,
    norm_out: Optional[torch.Tensor] = None,
    quant_out: Optional[torch.Tensor] = None,
    scale_out: Optional[torch.Tensor] = None,
    scale_factor: Optional[torch.Tensor] = None,
) -> None:
    num_tokens, hidden_size = allreduce_in.shape
    element_size = allreduce_in.element_size()
    current_tensor_size = num_tokens * hidden_size * element_size
    max_fusion_size = max_token_num * hidden_size * element_size
    use_flashinfer = current_tensor_size <= min(
        _FI_MAX_SIZES.get(world_size, _DEFAULT_FI_MAX_SIZE),
        max_fusion_size,
    )
    if use_flashinfer:
        assert (_FI_WORKSPACE_TENSOR is not None
                ), "Flashinfer must be enabled when using flashinfer"
        if norm_out is None:
            norm_out = allreduce_in
            residual_out = residual
        else:
            # return residual_out as allreduce_out with zeroed residual_in
            # as flashinfer does not support rms_norm
            # and allreduce_out together
            residual_out = allreduce_in
        # For the sizes that are smaller than the max size,
        # we only use flashinfer one shot allreduce
        flashinfer_comm.trtllm_allreduce_fusion(
            allreduce_in=allreduce_in,
            token_num=allreduce_in.shape[0],
            residual_in=residual,
            residual_out=residual_out,
            norm_out=norm_out,
            rms_gamma=rms_gamma,
            rms_eps=rms_eps,
            world_rank=world_rank,
            world_size=world_size,
            hidden_dim=allreduce_in.shape[-1],
            workspace_ptrs=_FI_WORKSPACE_TENSOR,
            launch_with_pdl=launch_with_pdl,
            use_oneshot=True,
            trigger_completion_at_end=trigger_completion_at_end,
            fp32_acc=fp32_acc,
            pattern_code=pattern_code,
            allreduce_out=None,
            quant_out=quant_out,
            scale_out=scale_out,
            # in vllm we only support swizzled layout
            layout_code=flashinfer_comm.QuantizationSFLayout.
            SWIZZLED_128x4,
            scale_factor=scale_factor,
        )
    else:
        allreduce_out = tensor_model_parallel_all_reduce(allreduce_in)
        if (scale_factor is not None and scale_out is None
                and fuse_rms_quant):
            # Do fused rms norm static fp8 quant fused op
            if norm_out is None:
                torch.ops._C.fused_add_rms_norm_static_fp8_quant(
                    quant_out, allreduce_out, residual, rms_gamma,
                    scale_factor, rms_eps)
            else:
                torch.ops._C.rms_norm_static_fp8_quant(
                    quant_out, allreduce_out, rms_gamma, scale_factor,
                    rms_eps)
        else:
            if norm_out is None:
                torch.ops._C.fused_add_rms_norm(allreduce_out, residual,
                                                rms_gamma, rms_eps)
                norm_out = allreduce_out
            else:
                torch.ops._C.rms_norm(norm_out, allreduce_out, rms_gamma,
                                      rms_eps)
            if scale_factor is not None:
                if scale_out is not None:
                    torch.ops._C.scaled_fp4_quant(quant_out, norm_out,
                                                  scale_out, scale_factor)
                else:
                    torch.ops._C.static_scaled_fp8_quant(
                        quant_out, norm_out, scale_factor)
        if scale_factor is None or norm_out is not None:
            # we need to return allreduce outpput
            # in cases of non quant fused AR + RMS norm
            # and fused AR + RMS norm + quant without fused add
            allreduce_in.copy_(allreduce_out)

call_trtllm_fused_allreduce_norm_fake 

call_trtllm_fused_allreduce_norm_fake(
    allreduce_in: Tensor,
    residual: Tensor,
    rms_gamma: Tensor,
    rms_eps: float,
    world_rank: int,
    world_size: int,
    launch_with_pdl: bool,
    trigger_completion_at_end: bool,
    fp32_acc: bool,
    max_token_num: int,
    pattern_code: int,
    fuse_rms_quant: bool,
    norm_out: Optional[Tensor] = None,
    quant_out: Optional[Tensor] = None,
    scale_out: Optional[Tensor] = None,
    scale_factor: Optional[Tensor] = None,
) -> None
Source code in vllm/compilation/collective_fusion.py 
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
def call_trtllm_fused_allreduce_norm_fake(
        allreduce_in: torch.Tensor,
        residual: torch.Tensor,
        rms_gamma: torch.Tensor,
        rms_eps: float,
        world_rank: int,
        world_size: int,
        launch_with_pdl: bool,
        trigger_completion_at_end: bool,
        fp32_acc: bool,
        max_token_num: int,
        pattern_code: int,
        fuse_rms_quant: bool,
        norm_out: Optional[torch.Tensor] = None,
        quant_out: Optional[torch.Tensor] = None,
        scale_out: Optional[torch.Tensor] = None,
        scale_factor: Optional[torch.Tensor] = None) -> None:
    pass