[TSP] Support qwen3 moe tsp + cudagraph (#4871)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
CE Compile Job / Show Code Archive Output (push) Has been cancelled
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Deploy GitHub Pages / deploy (push) Has been cancelled

* support qwen3_moe tsp mode

* fix

* fix

* update

* update

* update

* fix

* support external_rmsnorm

* update

* fix
This commit is contained in:
Yuanle Liu
2025-11-10 23:37:51 +08:00
committed by GitHub
parent fb2eb403ab
commit 3dc0ffa46d
28 changed files with 173 additions and 273 deletions

View File

@@ -360,6 +360,7 @@ export BKCL_RDMA_VERBS=1
export enable_expert_parallel=1
export enable_tensor_parallel=1
export disable_sequence_parallel_moe=1
python -m pytest -s --timeout=600 tests/ci_use/XPU_45T/run_ep.py
ep_exit_code=$?
@@ -373,6 +374,7 @@ unset XSHMEM_QP_NUM_PER_RANK
unset BKCL_RDMA_VERBS
unset enable_expert_parallel
unset enable_tensor_parallel
unset disable_sequence_parallel_moe
stop_processes
if [ ${ep_exit_code} -ne 0 ]; then
@@ -400,7 +402,6 @@ export BKCL_RDMA_VERBS=1
export enable_expert_parallel=1
export enable_tensor_parallel=1
export FD_EP_TP_STRATEGY=all_to_all
python -m pytest -s --timeout=600 tests/ci_use/XPU_45T/run_ep.py
ep_exit_code=$?
@@ -414,7 +415,6 @@ unset XSHMEM_QP_NUM_PER_RANK
unset BKCL_RDMA_VERBS
unset enable_expert_parallel
unset enable_tensor_parallel
unset FD_EP_TP_STRATEGY
stop_processes
if [ ${ep_exit_code} -ne 0 ]; then