Skip to content
Closed
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
49 changes: 27 additions & 22 deletions vllm/compilation/passes/pass_manager.py
Original file line number Diff line number Diff line change
Expand Up @@ -114,34 +114,39 @@ def configure(self, config: VllmConfig) -> None:
if self.pass_config.eliminate_noops:
self.passes += [NoOpEliminationPass(config)]

if self.pass_config.enable_sp:
self.passes += [SequenceParallelismPass(config)]
if self.pass_config.fuse_gemm_comms:
self.passes += [AsyncTPPass(config)]

if self.pass_config.fuse_allreduce_rms:
self.passes += [AllReduceFusionPass(config)]
if current_platform.is_cuda_alike():
if self.pass_config.enable_sp:
self.passes += [SequenceParallelismPass(config)]

if self.pass_config.fuse_norm_quant:
self.passes += [RMSNormQuantFusionPass(config)]
if rocm_aiter_ops.is_enabled():
self.passes += [
RocmAiterRMSNormQuantFusionPass(config),
]
if self.pass_config.fuse_act_quant:
self.passes += [ActivationQuantFusionPass(config)]
if rocm_aiter_ops.is_enabled():
self.passes += [RocmAiterSiluMulFp8GroupQuantFusionPass(config)]
# AsyncTPPass and AllReduceFusionPass are CUDA-only
if current_platform.is_cuda():
if self.pass_config.enable_sp and self.pass_config.fuse_gemm_comms:
self.passes += [AsyncTPPass(config)]
if self.pass_config.fuse_allreduce_rms:
self.passes += [AllReduceFusionPass(config)]

if current_platform.is_cuda_alike():
if self.pass_config.fuse_norm_quant:
self.passes += [RMSNormQuantFusionPass(config)]
if rocm_aiter_ops.is_enabled():
self.passes += [
RocmAiterRMSNormQuantFusionPass(config),
]
if self.pass_config.fuse_act_quant:
self.passes += [ActivationQuantFusionPass(config)]
if rocm_aiter_ops.is_enabled():
self.passes += [RocmAiterSiluMulFp8GroupQuantFusionPass(config)]

if self.pass_config.fuse_act_padding and rocm_aiter_ops.is_enabled():
self.passes += [RocmAiterTritonAddRMSNormPadFusionPass(config)]

if self.pass_config.fuse_attn_quant:
self.passes += [AttnFusionPass(config)]
if current_platform.is_cuda_alike():
if self.pass_config.fuse_attn_quant:
self.passes += [AttnFusionPass(config)]

if self.pass_config.enable_qk_norm_rope_fusion:
self.passes += [SplitCoalescingPass(config)]
self.passes += [QKNormRoPEFusionPass(config)]
if self.pass_config.enable_qk_norm_rope_fusion:
self.passes += [SplitCoalescingPass(config)]
self.passes += [QKNormRoPEFusionPass(config)]

# needs a functional graph
self.post_cleanup = PostCleanupPass(config)
Expand Down