Skip to content

[https://nvbugs/5448442][fix] Skip trtllm moe backend for sm120 #7010

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Aug 21, 2025
Merged

[https://nvbugs/5448442][fix] Skip trtllm moe backend for sm120 #7010

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8b4adfa
skip trtllm moe backend for sm120
pamelap-nvidia Aug 18, 2025
0565af1
add checks in trtllm gen fused moe module
pamelap-nvidia Aug 20, 2025
b857eae
Merge branch 'release/1.0' into fix_sm120_moe_trtllm
pamelap-nvidia Aug 20, 2025
03d315a
add checks in trtllm gen fused moe module
pamelap-nvidia Aug 20, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 24 additions & 2 deletions tests/integration/defs/accuracy/test_llm_api_pytorch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1328,6 +1328,11 @@ def test_nvfp4_4gpus_online_eplb(self, fp8kv):
@parametrize_with_ids("moe_backend", ["CUTLASS", "TRTLLM"])
def test_nvfp4(self, fp8kv, attention_dp, cuda_graph, overlap_scheduler,
torch_compile, mtp_nextn, moe_backend):
if moe_backend == "TRTLLM" and (get_sm_version() == 120
or get_sm_version() == 121):
pytest.skip(
"MOE TRTLLM backend does not support SM version 120 or 121")

kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
torch_compile_config = TorchCompileConfig(
enable_fullgraph=True,
Expand Down Expand Up @@ -1375,8 +1380,10 @@ def test_nvfp4_4gpus(self, fp8kv, attention_dp, cuda_graph,
torch_compile, mtp_nextn, moe_backend):
if torch_compile and pp_size > 1:
pytest.skip("PP with torch.compile is not supported yet.")
if moe_backend == "TRTLLM" and get_sm_version() == 120:
pytest.skip("MOE TRTLLM backend does not support SM version 120")
if moe_backend == "TRTLLM" and (get_sm_version() == 120
or get_sm_version() == 121):
pytest.skip(
"MOE TRTLLM backend does not support SM version 120 or 121")
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
# Picewise Cuda Graph cannot be enabled for nvfp4 attention dp.
torch_compile_config = TorchCompileConfig(
Expand Down Expand Up @@ -1591,6 +1598,11 @@ class TestDeepSeekR1(LlmapiAccuracyTestHarness):
def test_nvfp4_multi_gpus(self, tp_size, pp_size, ep_size, mtp_nextn, fp8kv,
attention_dp, cuda_graph, overlap_scheduler,
max_batch_size, moe_backend):
if moe_backend == "TRTLLM" and (get_sm_version() == 120
or get_sm_version() == 121):
pytest.skip(
"MOE TRTLLM backend does not support SM version 120 or 121")

kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.70)
pytorch_config = dict(
disable_overlap_scheduler=not overlap_scheduler,
Expand Down Expand Up @@ -2148,6 +2160,11 @@ def test_nvfp4(
torch_compile,
):

if moe_backend == "TRTLLM" and (get_sm_version() == 120
or get_sm_version() == 121):
pytest.skip(
"MOE TRTLLM backend does not support SM version 120 or 121")

torch_compile_config = TorchCompileConfig(
enable_fullgraph=True,
enable_piecewise_cuda_graph=cuda_graph and not attention_dp,
Expand Down Expand Up @@ -2268,6 +2285,11 @@ def test_fp8(self, tp_size, pp_size, ep_size, attention_dp, cuda_graph,
def test_nvfp4(self, tp_size, pp_size, ep_size, attention_dp, cuda_graph,
overlap_scheduler, moe_backend, eagle3):

if moe_backend == "TRTLLM" and (get_sm_version() == 120
or get_sm_version() == 121):
pytest.skip(
"MOE TRTLLM backend does not support SM version 120 or 121")

pytorch_config = dict(
disable_overlap_scheduler=not overlap_scheduler,
cuda_graph_config=CudaGraphConfig() if cuda_graph else None,
Expand Down
2 changes: 0 additions & 2 deletions tests/integration/test_lists/qa/llm_function_rtx6kd.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,6 @@ accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUT
accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_cutlass-torch_compile=False]
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_cutlass-torch_compile=True]
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_trtllm-torch_compile=False]
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_trtllm-torch_compile=True]
test_e2e.py::test_ptp_quickstart_advanced_mixed_precision
test_e2e.py::test_ptp_quickstart_advanced[Llama3.1-8B-BF16-llama-3.1-model/Meta-Llama-3.1-8B]
test_e2e.py::test_ptp_quickstart_advanced[Llama3.1-8B-FP8-llama-3.1-model/Llama-3.1-8B-Instruct-FP8]
Expand Down