biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Perf] Move eplb rebalance algo to async thread (#30888)

Browse Source 
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
This commit is contained in:
Ilya Markov
2026-02-10 23:19:10 +01:00
committed by GitHub
parent f0ca0671c7
commit 67132945bb
5 changed files with 251 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
tests/distributed/test_eplb_execute.py
View File

@@ -295,12 +295,11 @@ def _test_async_transfer_layer_without_mtp_worker(
for layer_idx in range(num_layers):
is_unchanged, is_received_locally, recv_metadata = asyncio.run(
transfer_layer(
old_global_expert_indices=old_indices_cpu,
new_global_expert_indices=new_indices_cpu,
expert_weights=expert_weights,
old_layer_indices=old_indices_cpu[layer_idx],
new_layer_indices=new_indices_cpu[layer_idx],
expert_weights=expert_weights[layer_idx],
expert_weights_buffer=expert_buffer,
ep_group=ep_group,
layer=layer_idx,
cuda_stream=cuda_stream,
)
)