[EPLB] Optimize EPLB with numpy

tests/distributed/test_eplb_algo.py

@@ -310,3 +310,136 @@ def test_additional_cases():
     print(phy2log)
 
     test_basic_rebalance()
+
+
+def _make_phyrank_from_phy2log(phy2log: torch.Tensor) -> torch.Tensor:
+    """Create phyrank from phy2log"""
+    pr = torch.zeros_like(phy2log)
+    for layer in range(phy2log.shape[0]):
+        seen: dict[int, int] = {}
+        row = phy2log[layer].tolist()
+        for i, expert in enumerate(row):
+            r = seen.get(expert, 0)
+            pr[layer, i] = r
+            seen[expert] = r + 1
+    return pr
+
+
+def _validate_intragpu_rearrangement(
+    old_global_expert_indices: torch.Tensor,
+    new_phy2log: torch.Tensor,
+    new_phyrank: torch.Tensor,
+    post_phy2log: torch.Tensor,
+    post_phyrank: torch.Tensor,
+    num_ranks: int,
+    slots_per_gpu: int,
+):
+    # Per-GPU checks
+    for gpu_idx in range(num_ranks):
+        start = gpu_idx * slots_per_gpu
+        end = start + slots_per_gpu
+        old_seg = old_global_expert_indices[0, start:end]
+        new_seg = new_phy2log[0, start:end]
+        new_rnk = new_phyrank[0, start:end]
+        post_seg = post_phy2log[0, start:end]
+        post_rnk = post_phyrank[0, start:end]
+
+        # Pairwise equality for (expert, rank) pairs to ensure nothing is lost
+        def sorted_pairs(seg: torch.Tensor, rnk: torch.Tensor):
+            pairs = list(zip(seg.tolist(), rnk.tolist()))
+            pairs.sort()
+            return pairs
+
+        assert sorted_pairs(post_seg, post_rnk) == sorted_pairs(new_seg, new_rnk), (
+            f"Per-GPU pairs of (expert,rank) must match new mapping for GPU {gpu_idx}"
+        )
+
+        # For experts that remain on the same GPU, the old slot is preserved
+        # for at least one occurrence; rank at that slot must be valid for that expert
+        old_list = old_seg.tolist()
+        new_list = new_seg.tolist()
+        post_list = post_seg.tolist()
+        remained = set(old_list) & set(new_list)
+        new_ranks_for_expert: dict[int, list[int]] = {}
+        for v, r in zip(new_list, new_rnk.tolist()):
+            new_ranks_for_expert.setdefault(v, []).append(r)
+        for expert in remained:
+            old_pos = old_list.index(expert)
+            assert post_list[old_pos] == expert, (
+                f"Expert {expert} on GPU {gpu_idx} should stay at old slot {old_pos}"
+            )
+            # Rank at preserved slot must be one of the ranks
+            # the expert has in new mapping
+            assert post_rnk.tolist()[old_pos] in new_ranks_for_expert[expert], (
+                f"Rank for expert {expert} at preserved slot on GPU {gpu_idx} "
+                "must come from new mapping"
+            )
+
+
+def test_preserve_intragpu_slots_simple():
+    """Experts that stay on a GPU keep their old slots; incoming not lost."""
+    # Setup: 2 GPUs, 4 slots each, 1 layer
+    num_ranks = 2
+    slots_per_gpu = 4
+    # Old mapping: GPU0 -> [0,1,2,3], GPU1 -> [4,5,6,7]
+    old_global_expert_indices = torch.tensor([[0, 1, 2, 3, 4, 5, 6, 7]])
+    # New mapping shuffles within GPU0 and brings 4,5 into GPU0.
+    # GPU0 new -> [1,5,0,4] (0 and 1 remain on GPU0 but at different slots)
+    # GPU1 new -> [6,2,7,3] (6 and 7 remain on GPU1, 2 and 3 move in)
+    phy2log = torch.tensor([[1, 5, 0, 4, 6, 2, 7, 3]])
+    # Derive phyrank from replica occurrence order per expert
+    phyrank = _make_phyrank_from_phy2log(phy2log)
+
+    post_phy2log, post_phyrank = DefaultEplbPolicy.preserve_intragpu_slots(
+        phy2log, phyrank, num_ranks, old_global_expert_indices
+    )
+
+    # Shapes preserved
+    assert post_phy2log.shape == phy2log.shape
+    assert post_phyrank.shape == phyrank.shape
+
+    _validate_intragpu_rearrangement(
+        old_global_expert_indices,
+        phy2log,
+        phyrank,
+        post_phy2log,
+        post_phyrank,
+        num_ranks,
+        slots_per_gpu,
+    )
+
+
+def test_preserve_intragpu_slots_with_duplicates():
+    """Test preserve intragpu slots with duplicates"""
+    # Setup: 2 GPUs, 5 slots each (total 10 physical experts), 1 layer
+    num_ranks = 2
+    slots_per_gpu = 5
+    # Old mapping:
+    #   GPU0 -> [0, 1, 0, 2, 3]  (expert 0 duplicated)
+    #   GPU1 -> [4, 5, 6, 1, 2]
+    old_global_expert_indices = torch.tensor([[0, 1, 0, 2, 3, 4, 5, 6, 1, 2]])
+    # New mapping reorders within GPUs and moves some experts across GPUs,
+    # while still including duplicates:
+    #   GPU0 new -> [0, 5, 4, 0, 1]  (expert 0 duplicated, 4/5 incoming)
+    #   GPU1 new -> [6, 2, 3, 1, 2]  (expert 2 duplicated)
+    phy2log = torch.tensor([[0, 5, 4, 0, 1, 6, 2, 3, 1, 2]])
+    # Derive ranks so duplicates have ranks [0,1,...] by occurrence
+    phyrank = _make_phyrank_from_phy2log(phy2log)
+
+    post_phy2log, post_phyrank = DefaultEplbPolicy.preserve_intragpu_slots(
+        phy2log, phyrank, num_ranks, old_global_expert_indices
+    )
+
+    # Shapes preserved
+    assert post_phy2log.shape == phy2log.shape
+    assert post_phyrank.shape == phyrank.shape
+
+    _validate_intragpu_rearrangement(
+        old_global_expert_indices,
+        phy2log,
+        phyrank,
+        post_phy2log,
+        post_phyrank,
+        num_ranks,
+        slots_per_gpu,
+    )

tests/distributed/test_eplb_execute.py

@@ -286,31 +286,32 @@ def _test_async_transfer_layer_without_mtp_worker(
         device,
         old_indices,
     )
+    old_indices_cpu = old_indices.cpu()
+    new_indices_cpu = new_indices.cpu()
 
     expert_buffer = [torch.empty_like(w) for w in expert_weights[0]]
     cuda_stream = torch.cuda.Stream(device=device)
 
     for layer_idx in range(num_layers):
-        is_unchanged, is_received_locally, experts_recv_loc = asyncio.run(
+        is_unchanged, is_received_locally, recv_metadata = asyncio.run(
             transfer_layer(
-                old_global_expert_indices=old_indices,
-                new_global_expert_indices=new_indices,
+                old_global_expert_indices=old_indices_cpu,
+                new_global_expert_indices=new_indices_cpu,
                 expert_weights=expert_weights,
                 expert_weights_buffer=expert_buffer,
                 ep_group=ep_group,
                 layer=layer_idx,
                 cuda_stream=cuda_stream,
             )
         )
-
         cuda_stream.synchronize()
         move_from_buffer(
             expert_weights=expert_weights[layer_idx],
-            expert_weights_buffer=expert_buffer,
+            expert_weights_buffers=expert_buffer,
             is_unchanged=is_unchanged,
             is_received_locally=is_received_locally,
-            experts_recv_loc=experts_recv_loc,
-            new_indices=new_indices[layer_idx].tolist(),
+            recv_metadata=recv_metadata,
+            new_indices=new_indices_cpu[layer_idx],
             ep_group=ep_group,
         )
 

vllm/config/parallel.py

@@ -61,6 +61,10 @@ class EPLBConfig:
     Log the balancedness each step of expert parallelism.
     This is turned off by default since it will cause communication overhead.
     """
+    log_balancedness_interval: int = 1
+    """
+    Interval for logging the balancedness.
+    """
     use_async: bool = False
     """
     Whether to use non-blocking EPLB.

vllm/distributed/eplb/async_worker.py

@@ -89,7 +89,7 @@ async def transfer_run_periodically(
                         (
                             model_state.is_unchanged,
                             model_state.is_received_locally,
-                            model_state.experts_recv_loc,
+                            model_state.recv_metadata,
                         ) = await transfer_layer(
                             old_global_expert_indices=model_state.physical_to_logical_map,
                             new_global_expert_indices=model_state.new_physical_to_logical_map,

vllm/distributed/eplb/eplb_state.py

@@ -27,10 +27,10 @@
 """
 
 import threading
-import time
 from collections.abc import Sequence
 from dataclasses import dataclass
 
+import numpy as np
 import torch
 from torch.distributed import ProcessGroup, all_reduce
 
@@ -46,7 +46,11 @@
 
 from .async_worker import start_async_worker
 from .policy import EPLB_POLICIES, AbstractEplbPolicy, DefaultEplbPolicy
-from .rebalance_execute import move_from_buffer, rearrange_expert_weights_inplace
+from .rebalance_execute import (
+    RecvMetadata,
+    move_from_buffer,
+    rearrange_expert_weights_inplace,
+)
 
 logger = init_logger(__name__)
 
@@ -164,20 +168,19 @@ class EplbModelState:
     """
     Whether the async EPLB needs to poll peers for buffer readiness.
     """
-    is_unchanged: list[bool]
+    is_unchanged: np.ndarray
     """
     intermediate variable between `move_to_buffer` and `move_to_workspace`.
     The size is same as the num of physical experts in the current layer.
     """
-    is_received_locally: list[bool]
+    is_received_locally: np.ndarray
     """
     intermediate variable between `move_to_buffer` and `move_to_workspace`.
     The size is same as the num of physical experts in the current layer.
     """
-    experts_recv_loc: dict[int, int]
+    recv_metadata: RecvMetadata
     """
     intermediate variable between `move_to_buffer` and `move_to_workspace`.
-    The size is same as the num of physical experts in the current layer.
     """
     is_async_enabled: bool
     """
@@ -507,9 +510,14 @@ def add_model(
             layer_to_transfer=0,
             rebalanced=False,
             pending_global_ready_check=False,
-            is_unchanged=[],
-            is_received_locally=[],
-            experts_recv_loc={},
+            is_unchanged=np.array([]),
+            is_received_locally=np.array([]),
+            recv_metadata=RecvMetadata(
+                recv_primary_mask=np.array([]),
+                recv_count=0,
+                recv_expert_ids=np.array([]),
+                recv_dst_rows=np.array([]),
+            ),
             is_async_enabled=self.is_async,
             cuda_device_index=self.cuda_device_index,
             new_physical_to_logical_map=new_physical_to_logical_map,
@@ -553,7 +561,12 @@ def step(
             for eplb_model_state in self.model_states.values():
                 eplb_model_state.expert_load_pass.zero_()
 
-        if log_stats:
+        if (
+            log_stats
+            and self.expert_rearrangement_step
+            % self.parallel_config.eplb_config.log_balancedness_interval
+            == 0
+        ):
             # Sync the expert load pass for each model (main and drafter).
             # expert_load_pass: (num_moe_layers, num_physical_experts)
             expert_load_pass_list = self._sync_load_pass()
@@ -585,9 +598,10 @@ def step(
 
                 if ep_group.rank() == 0:
                     logger.info(
-                        "EPLB step: %d for model %s: avg_tokens=%.2f, "
+                        "EPLB step: %d/%d for model %s: avg_tokens=%.2f, "
                         "max_tokens=%d, balancedness=%.4f",
                         self.expert_rearrangement_step,
+                        self.expert_rearrangement_step_interval,
                         eplb_model_state.model_name,
                         avg_tokens,
                         max_tokens,
@@ -684,11 +698,14 @@ def rearrange(
         ep_group = get_ep_group().device_group
         ep_rank = ep_group.rank()
 
-        time_start = None
+        start_event = None
+        end_event = None
         is_main_rank = ep_rank == 0
         if is_main_rank:
-            torch.cuda.synchronize()
-            time_start = time.perf_counter()
+            if not self.is_async or is_profile:
+                start_event = torch.cuda.Event(enable_timing=True)
+                end_event = torch.cuda.Event(enable_timing=True)
+                start_event.record()
             logger.info(
                 "Rearranging experts %s %s...",
                 "(async mode)" if self.is_async else "sync mode",
@@ -800,6 +817,7 @@ def rearrange(
                 num_groups,
                 num_nodes,
                 num_gpus,
+                eplb_model_state.physical_to_logical_map,
             )
 
             if not eplb_model_state.is_async_enabled or is_profile:
@@ -848,17 +866,17 @@ def rearrange(
                         new_logical_replica_count
                     )
                 if is_main_rank:
-                    assert time_start is not None
-                    torch.cuda.synchronize()
-                    time_end = time.perf_counter()
+                    assert start_event is not None
+                    assert end_event is not None
+                    end_event.record()
+                    end_event.synchronize()
+                    gpu_elapsed = start_event.elapsed_time(end_event) / 1000.0
                     logger.info(
-                        "Rearranged experts%sin %.2f seconds.",
+                        "Rearranged experts %s in %.2f s.",
                         " (profile) " if is_profile else " ",
-                        time_end - time_start,
+                        gpu_elapsed,
                     )
             else:
-                device = eplb_model_state.physical_to_logical_map.device
-                new_physical = new_physical_to_logical_map.to(device)
                 max_slots = eplb_model_state.logical_to_physical_map.shape[-1]
                 padded_logical = torch.nn.functional.pad(
                     new_logical_to_physical_map,
@@ -869,7 +887,10 @@ def rearrange(
                     eplb_model_state.logical_replica_count.device
                 )
 
-                eplb_model_state.new_physical_to_logical_map = new_physical
+                # Move map to cpu in advance
+                eplb_model_state.new_physical_to_logical_map = (
+                    new_physical_to_logical_map.cpu()
+                )
                 eplb_model_state.new_logical_to_physical_map = padded_logical
                 eplb_model_state.new_logical_replica_count = new_replica
 
@@ -968,25 +989,30 @@ def move_to_workspace(
                 stream = torch.cuda.current_stream(device=device_index)
                 stream.wait_event(model_state.buffer_ready_event)
                 model_state.buffer_ready_event = None
+            expert_weights = model_state.model.expert_weights[
+                model_state.layer_to_transfer
+            ]
+            expert_weights_buffer = model_state.expert_buffer
+            new_indices = (
+                model_state.new_physical_to_logical_map[model_state.layer_to_transfer]
+                .cpu()
+                .numpy()
+            )
             move_from_buffer(
-                expert_weights=model_state.model.expert_weights[
-                    model_state.layer_to_transfer
-                ],
-                expert_weights_buffer=model_state.expert_buffer,
+                expert_weights=expert_weights,
+                expert_weights_buffers=expert_weights_buffer,
                 is_unchanged=model_state.is_unchanged,
                 is_received_locally=model_state.is_received_locally,
-                experts_recv_loc=model_state.experts_recv_loc,
-                new_indices=model_state.new_physical_to_logical_map[
-                    model_state.layer_to_transfer
-                ].tolist(),
-                ep_group=ep_group,
+                recv_metadata=model_state.recv_metadata,
+                new_indices=new_indices,
+                ep_rank=ep_group.rank(),
             )
             transferred_layer = model_state.layer_to_transfer
             self._update_layer_mapping_from_new(model_state, transferred_layer)
             # After the main thread consumes, advance layer_to_transfer
             model_state.layer_to_transfer += 1
             model_state.ep_buffer_ready = 0
-            logger.info(
+            logger.debug(
                 "model %s successfully move_to_workspace layer %d",
                 model_state.model_name,
                 transferred_layer,