fix ds-v2 run error

vllm/attention/ops/flashmla.py

@@ -103,6 +103,7 @@ def flash_mla_with_kvcache(
             out, softmax_lse = flash_mla_cuda.fwd_kvcache_mla(
                 q,
                 k_cache,
+                None,
                 head_dim_v,
                 cache_seqlens,
                 block_table,
@@ -117,6 +118,7 @@ def flash_mla_with_kvcache(
         out, softmax_lse = flash_mla_cuda.fwd_kvcache_mla(
             q,
             k_cache,
+            None,
             head_dim_v,
             cache_seqlens,
             block_table,
@@ -129,6 +131,7 @@ def flash_mla_with_kvcache(
         out, softmax_lse = torch.ops._flashmla_C.fwd_kvcache_mla(
             q,
             k_cache,
+            None,
             head_dim_v,
             cache_seqlens,
             block_table,

vllm/model_executor/layers/fused_moe/fused_moe.py

@@ -884,7 +884,6 @@ def invoke_fused_moe_kernel(A: torch.Tensor,
             use_int8_w8a16=use_int8_w8a16,
             per_channel_quant=per_channel_quant,
             HAS_BIAS=HAS_BIAS,
-            BLOCK_SIZE_K=BLOCK_SIZE_K,
             **config,
         )
 

vllm/v1/attention/backends/mla/common.py

@@ -490,7 +490,6 @@ class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
         self.block_table = block_table
         
         self.use_spec_decode = False
-        self.num_scheduled_tokens_np = np.zeros(scheduler_config.max_num_seqs, dtype=np.int32)
         
         # support for cudagraph spec docoding
         self.spec_decode_block_table_tensor = None

vllm/v1/attention/backends/utils.py

@@ -18,6 +18,7 @@ if TYPE_CHECKING:
     from vllm.attention.backends.abstract import AttentionImpl
     from vllm.v1.core.sched.output import SchedulerOutput
     from vllm.v1.worker.gpu_input_batch import InputBatch
+    from vllm.v1.attention.backends.mla.common import MLACommonMetadataBuilder
 
 import vllm.envs as envs
 from vllm.attention.backends.abstract import AttentionBackend
@@ -630,7 +631,7 @@ def reorder_batch_to_split_decodes_and_prefills(
     input_batch: "InputBatch",
     scheduler_output: "SchedulerOutput",
     decode_threshold: int = 1,
-    num_scheduled_tokens_np: Optional[np.ndarray] = None
+    # num_scheduled_tokens_np: np.ndarray = np.zeros(256, dtype=np.int32),
 ) -> bool:
     """
     Reorders the batch to split into prefill and decode requests; places all
@@ -650,6 +651,9 @@ def reorder_batch_to_split_decodes_and_prefills(
     num_decode_tokens = 0
     num_prefill_tokens = 0
     
+    vllm_config = VllmConfig()
+    num_scheduled_tokens_np = np.zeros(vllm_config.scheduler_config.max_num_seqs, dtype=np.int32)
+
     for i, req_id in enumerate(input_batch.req_ids):
         num_tokens = scheduler_output.num_scheduled_tokens[req_id]
         # for now treat 1 scheduled token as "decode" even if its not,