[Feature] PD-Multiplexing Context and Scheduler. (#11592)

python/sglang/srt/managers/schedule_batch.py

@@ -70,7 +70,11 @@ from sglang.srt.mem_cache.memory_pool import ReqToTokenPool
 from sglang.srt.mem_cache.radix_cache import RadixKey
 from sglang.srt.mem_cache.swa_radix_cache import SWARadixCache
 from sglang.srt.metrics.collector import SchedulerMetricsCollector, TimeStats
-from sglang.srt.model_executor.forward_batch_info import CaptureHiddenMode, ForwardMode
+from sglang.srt.model_executor.forward_batch_info import (
+    CaptureHiddenMode,
+    ForwardBatch,
+    ForwardMode,
+)
 from sglang.srt.sampling.sampling_batch_info import SamplingBatchInfo
 from sglang.srt.sampling.sampling_params import SamplingParams
 from sglang.srt.server_args import ServerArgs, get_global_server_args
@@ -1010,6 +1014,13 @@ class ScheduleBatch(ScheduleBatchDisaggregationDecodeMixin):
     encoder_lens_cpu: Optional[List[int]] = None
     encoder_out_cache_loc: Optional[torch.Tensor] = None
 
+    # For split prefill
+    split_index: int = 0
+    split_prefill_finished: bool = False
+    split_forward_count: int = 1
+    split_forward_batch: ForwardBatch = None
+    seq_lens_cpu_cache: torch.Tensor = None
+
     # Stream
     has_stream: bool = False