[ROCm][FEAT] Fuse DeepSeek shared experts into AITER fused_moe ops (#24097)
Signed-off-by: chenjun <junchen2@amd.com> Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com> Co-authored-by: valarLip <103567126+valarLip@users.noreply.github.com> Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
This commit is contained in:
@@ -217,7 +217,7 @@ def test_moe_permute_unpermute(
|
||||
expert_map = None
|
||||
n_local_expert = n_expert
|
||||
if ep_size != 1:
|
||||
n_local_expert, expert_map = determine_expert_map(ep_size, ep_rank, n_expert)
|
||||
n_local_expert, expert_map, _ = determine_expert_map(ep_size, ep_rank, n_expert)
|
||||
expert_map = expert_map.cuda()
|
||||
start_expert = n_local_expert * ep_rank
|
||||
current_platform.seed_everything(0)
|
||||
|
||||
Reference in New Issue
Block a user