[PCG] fix piecewise cuda graph for Qwen3.5 (#19220)

python/sglang/srt/layers/quantization/fp8_utils.py

@@ -72,6 +72,13 @@ if _is_cuda:
         N = mat_b.shape[-1]
         return mat_a.new_empty((M, N), dtype=out_dtype)
 
+    @torch.library.register_fake("sgl_kernel::fp8_blockwise_scaled_mm")
+    def _fp8_blockwise_scaled_mm_abstract(mat_a, mat_b, scales_a, scales_b, out_dtype):
+        # mat_a: [M, K], mat_b: [K, N] or [N, K] depending on callsite layout; output is [M, N].
+        M = mat_a.shape[-2]
+        N = mat_b.shape[-1]
+        return mat_a.new_empty((M, N), dtype=out_dtype)
+
 
 use_vllm_cutlass_w8a8_fp8_kernel = get_bool_env_var("USE_VLLM_CUTLASS_W8A8_FP8_KERNEL")
 use_triton_w8a8_fp8_kernel = get_bool_env_var("USE_TRITON_W8A8_FP8_KERNEL")

python/sglang/srt/models/qwen3_5.py

@@ -22,9 +22,6 @@ import torch
 import torch.nn as nn
 from einops import rearrange
 
-# Model Executor
-from sglang.srt.compilation.piecewise_context_manager import get_forward_context
-
 # Configs
 from sglang.srt.configs.qwen3_5 import (
     Qwen3_5Config,
@@ -72,7 +69,6 @@ from sglang.srt.model_loader.weight_utils import (
 from sglang.srt.models.qwen2_moe import Qwen2MoeMLP, Qwen2MoeSparseMoeBlock
 
 # Models
-from sglang.srt.models.qwen3_next import gdn_with_output
 from sglang.srt.models.qwen3_vl import Qwen3VLForConditionalGeneration
 
 # Utils
@@ -253,22 +249,6 @@ class Qwen3_5GatedDeltaNet(nn.Module):
         self,
         hidden_states: torch.Tensor,
         forward_batch: ForwardBatch,
-    ):
-        output = torch.empty_like(hidden_states)
-        if forward_batch.forward_mode.is_extend() and get_forward_context() is not None:
-            gdn_with_output(
-                hidden_states,
-                output,
-                self.layer_id,
-            )
-            return output
-        else:
-            return self._forward(hidden_states, forward_batch)
-
-    def _forward(
-        self,
-        hidden_states: torch.Tensor,
-        forward_batch: ForwardBatch,
     ):
         """
         Forward pass with three parts:
@@ -287,7 +267,7 @@ class Qwen3_5GatedDeltaNet(nn.Module):
         b = b.contiguous()
         a = a.contiguous()
 
-        core_attn_out = self.attn.forward(
+        core_attn_out = self.attn(
             forward_batch=forward_batch,
             mixed_qkv=mixed_qkv,
             a=a,

python/sglang/srt/models/qwen3_next.py

@@ -5,8 +5,6 @@ from typing import Any, Iterable, Optional, Set, Tuple
 import torch
 from torch import nn
 
-from sglang.srt.compilation.compilation_config import register_split_op
-from sglang.srt.compilation.piecewise_context_manager import get_forward_context
 from sglang.srt.configs.qwen3_next import Qwen3NextConfig
 from sglang.srt.distributed import get_pp_group
 from sglang.srt.eplb.expert_distribution import get_global_expert_distribution_recorder
@@ -53,7 +51,6 @@ from sglang.srt.utils import (
     make_layers,
     set_weight_attrs,
 )
-from sglang.srt.utils.custom_op import register_custom_op
 
 logger = logging.getLogger(__name__)
 _is_cuda = is_cuda()
@@ -1149,25 +1146,3 @@ class Qwen3NextForCausalLM(nn.Module):
 
 
 EntryClass = Qwen3NextForCausalLM
-
-
-@register_custom_op(mutates_args=["output"])
-@register_split_op()
-def gdn_with_output(
-    hidden_states: torch.Tensor,
-    output: torch.Tensor,
-    layer_id: int,
-) -> None:
-    context = get_forward_context()
-    forward_batch = context.forward_batch
-    attention_layers = context.attention_layers
-    attention_layer = attention_layers[layer_id]
-
-    ret = attention_layer._forward(hidden_states, forward_batch)
-
-    assert (
-        output.numel() == ret.numel()
-    ), f"Output tensor element mismatch: {output.numel()} != {ret.numel()}"
-
-    output.view(ret.shape).copy_(ret)
-    return

python/sglang/srt/models/qwen3_vl.py

@@ -1233,6 +1233,7 @@ class Qwen3VLForConditionalGeneration(nn.Module):
     def should_apply_lora(self, module_name: str) -> bool:
         return bool(self._lora_pattern.match(module_name))
 
+    @torch.no_grad()
     def forward(
         self,
         input_ids: torch.Tensor,