Skip to content

Commit b08dda3

Browse files
Wanli-Jiangmikeiovine
Wanli-Jiang
authored and
mikeiovine
committed
[None][fix] Bypass key-word matching for multimodal tests (NVIDIA#9170)
Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
1 parent 3fbbd44 commit b08dda3

File tree

8 files changed

+79
-74
lines changed

8 files changed

+79
-74
lines changed

‎tests/integration/defs/accuracy/references/mmmu.yaml‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -15,3 +15,5 @@ Efficient-Large-Model/VILA1.5-3b:
1515
# the metric here is for model sanity checking.
1616
nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16:
1717
- accuracy: 26.67
18+
microsoft/Phi-4-multimodal-instruct:
19+
- accuracy: 53.67

‎tests/integration/defs/accuracy/test_llm_api_pytorch.py‎

Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -4399,3 +4399,23 @@ def test_nvfp4_4gpus(self):
43994399
if temp_dir and os.path.exists(temp_dir):
44004400
import shutil
44014401
shutil.rmtree(temp_dir, ignore_errors=True)
4402+
4403+
4404+
class TestPhi4MMFusedVisionLora(LlmapiAccuracyTestHarness):
4405+
MODEL_NAME = "microsoft/Phi-4-multimodal-instruct"
4406+
MODEL_PATH = f"{llm_models_root()}/multimodals/Phi-4-multimodal-instruct-fuse-vision-lora"
4407+
MAX_NUM_TOKENS = 25600
4408+
4409+
sampling_params = SamplingParams(max_tokens=MAX_NUM_TOKENS,
4410+
truncate_prompt_tokens=MMMU.MAX_INPUT_LEN,
4411+
stop="<|USER|>")
4412+
4413+
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.7)
4414+
4415+
def test_auto_dtype(self):
4416+
with LLM(self.MODEL_PATH,
4417+
max_batch_size=32,
4418+
max_num_tokens=self.MAX_NUM_TOKENS,
4419+
kv_cache_config=self.kv_cache_config) as llm:
4420+
task = MMMU(self.MODEL_NAME)
4421+
task.evaluate(llm, sampling_params=self.sampling_params)

‎tests/integration/defs/test_e2e.py‎

Lines changed: 40 additions & 57 deletions
Original file line numberDiff line numberDiff line change
@@ -2556,7 +2556,9 @@ def test_ptp_quickstart_multimodal(llm_root, llm_venv, model_name, model_path,
25562556
print("output:", output)
25572557
return
25582558

2559-
match_ratio = 4.0 / 5
2559+
# Set match ratio to 0.0 to bypass keyword matching.
2560+
match_ratio = 0.0
2561+
25602562
parsed_outputs = parse_output(output)
25612563
for prompt_output, prompt_keywords in zip(
25622564
parsed_outputs, expected_keywords[model_name][modality]):
@@ -2570,29 +2572,21 @@ def test_ptp_quickstart_multimodal(llm_root, llm_venv, model_name, model_path,
25702572

25712573

25722574
@pytest.mark.parametrize("modality", ["image", "video"])
2573-
@pytest.mark.parametrize(
2574-
"model_name,model_path,match_ratio",
2575-
[
2576-
("phi4-multimodal-instruct", "multimodals/Phi-4-multimodal-instruct",
2577-
0.8),
2578-
pytest.param("phi4-multimodal-instruct-fp4",
2579-
"multimodals/Phi-4-multimodal-instruct-FP4",
2580-
0.8,
2581-
marks=skip_pre_blackwell),
2582-
pytest.param("phi4-multimodal-instruct-fp8",
2583-
"multimodals/Phi-4-multimodal-instruct-FP8",
2584-
0.8,
2585-
marks=skip_pre_hopper),
2586-
pytest.param(
2587-
"mistral-small-3.1-24b-instruct",
2588-
"Mistral-Small-3.1-24B-Instruct-2503",
2589-
# Lower threshold to give some wiggle room for flakiness.
2590-
0.6,
2591-
marks=pytest.mark.skip_less_device_memory(80000)),
2592-
])
2575+
@pytest.mark.parametrize("model_name,model_path", [
2576+
("phi4-multimodal-instruct", "multimodals/Phi-4-multimodal-instruct"),
2577+
pytest.param("phi4-multimodal-instruct-fp4",
2578+
"multimodals/Phi-4-multimodal-instruct-FP4",
2579+
marks=skip_pre_blackwell),
2580+
pytest.param("phi4-multimodal-instruct-fp8",
2581+
"multimodals/Phi-4-multimodal-instruct-FP8",
2582+
marks=skip_pre_hopper),
2583+
pytest.param("mistral-small-3.1-24b-instruct",
2584+
"Mistral-Small-3.1-24B-Instruct-2503",
2585+
marks=pytest.mark.skip_less_device_memory(80000)),
2586+
])
25932587
def test_ptp_quickstart_multimodal_kv_cache_reuse(llm_root, llm_venv,
25942588
model_name, model_path,
2595-
modality, match_ratio):
2589+
modality):
25962590
# NOTE: individual tests need to be enabled in
25972591
# tests/integration/test_lists/qa/examples_test_list.txt
25982592

@@ -2682,7 +2676,9 @@ def test_ptp_quickstart_multimodal_kv_cache_reuse(llm_root, llm_venv,
26822676
cmd.append("Phi4MMForCausalLM")
26832677

26842678
output = llm_venv.run_cmd(cmd, caller=check_output)
2685-
match_ratio = 4.0 / 5
2679+
2680+
# Set match ratio to 0.0 to bypass keyword matching.
2681+
match_ratio = 0.0
26862682
for prompt_output, prompt_keywords in zip(
26872683
parse_output(output), expected_keywords[model_name][modality]):
26882684
matches = [
@@ -2700,29 +2696,21 @@ def test_ptp_quickstart_multimodal_kv_cache_reuse(llm_root, llm_venv,
27002696

27012697

27022698
@pytest.mark.parametrize("modality", ["image", "video"])
2703-
@pytest.mark.parametrize(
2704-
"model_name,model_path,match_ratio",
2705-
[
2706-
("phi4-multimodal-instruct", "multimodals/Phi-4-multimodal-instruct",
2707-
0.8),
2708-
pytest.param("phi4-multimodal-instruct-fp4",
2709-
"multimodals/Phi-4-multimodal-instruct-FP4",
2710-
0.8,
2711-
marks=skip_pre_blackwell),
2712-
pytest.param("phi4-multimodal-instruct-fp8",
2713-
"multimodals/Phi-4-multimodal-instruct-FP8",
2714-
0.8,
2715-
marks=skip_pre_hopper),
2716-
pytest.param(
2717-
"mistral-small-3.1-24b-instruct",
2718-
"Mistral-Small-3.1-24B-Instruct-2503",
2719-
# Lower threshold to give some wiggle room for flakiness.
2720-
0.6,
2721-
marks=pytest.mark.skip_less_device_memory(80000)),
2722-
])
2699+
@pytest.mark.parametrize("model_name,model_path", [
2700+
("phi4-multimodal-instruct", "multimodals/Phi-4-multimodal-instruct"),
2701+
pytest.param("phi4-multimodal-instruct-fp4",
2702+
"multimodals/Phi-4-multimodal-instruct-FP4",
2703+
marks=skip_pre_blackwell),
2704+
pytest.param("phi4-multimodal-instruct-fp8",
2705+
"multimodals/Phi-4-multimodal-instruct-FP8",
2706+
marks=skip_pre_hopper),
2707+
pytest.param("mistral-small-3.1-24b-instruct",
2708+
"Mistral-Small-3.1-24B-Instruct-2503",
2709+
marks=pytest.mark.skip_less_device_memory(80000)),
2710+
])
27232711
def test_ptp_quickstart_multimodal_chunked_prefill(llm_root, llm_venv,
27242712
model_name, model_path,
2725-
modality, match_ratio):
2713+
modality):
27262714
# NOTE: individual tests need to be enabled in
27272715
# tests/integration/test_lists/qa/examples_test_list.txt
27282716

@@ -2841,6 +2829,8 @@ def test_ptp_quickstart_multimodal_chunked_prefill(llm_root, llm_venv,
28412829
cmd.append("Phi4MMForCausalLM")
28422830

28432831
output = llm_venv.run_cmd(cmd, caller=check_output)
2832+
# Set match ratio to 0.0 to bypass keyword matching.
2833+
match_ratio = 0.0
28442834
for prompt_output, prompt_keywords in zip(
28452835
parse_output(output), expected_keywords[model_name][modality]):
28462836
matches = [
@@ -2942,7 +2932,8 @@ def test_ptp_quickstart_multimodal_phi4mm(llm_root, llm_venv, model_name,
29422932
]
29432933
output = llm_venv.run_cmd(cmd, caller=check_output)
29442934

2945-
match_ratio = 0.6
2935+
# Set match ratio to 0.0 to bypass keyword matching.
2936+
match_ratio = 0.0
29462937
parsed_outputs = parse_output(output)
29472938
for prompt_output, prompt_keywords in zip(parsed_outputs,
29482939
expected_keywords[modality]):
@@ -3067,12 +3058,8 @@ def test_ptp_quickstart_multimodal_2gpu(llm_root, llm_venv, model_name,
30673058
print("output:", output)
30683059
return
30693060

3070-
# Set match ratio based on model
3071-
match_ratio = 4.0 / 5
3072-
if model_name.startswith("phi4-multimodal-instruct"):
3073-
match_ratio = 0.6
3074-
3075-
# Check output accuracy
3061+
# Set match ratio to 0.0 to bypass keyword matching.
3062+
match_ratio = 0.0
30763063
parsed_outputs = parse_output(output)
30773064
for prompt_output, prompt_keywords in zip(
30783065
parsed_outputs, expected_keywords[model_name]["image"]):
@@ -3195,12 +3182,8 @@ def test_ptp_quickstart_multimodal_multiturn(llm_root, llm_venv, model_name,
31953182
)
31963183
return
31973184

3198-
# Set match ratio based on model
3199-
match_ratio = 4.0 / 5
3200-
if model_name.startswith("Phi-4-multimodal-instruct"):
3201-
match_ratio = 0.6
3202-
3203-
# Check output accuracy
3185+
# Set match ratio to 0.0 to bypass keyword matching.
3186+
match_ratio = 0.0
32043187
parsed_outputs = parse_output(output)
32053188
for prompt_output, prompt_keywords in zip(
32063189
parsed_outputs, expected_keywords[model_name]["image"]):

‎tests/integration/test_lists/qa/llm_function_core.txt‎

Lines changed: 9 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -615,6 +615,7 @@ accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype
615615
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype_long_rope
616616
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp4
617617
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp8
618+
accuracy/test_llm_api_pytorch.py::TestPhi4MMFusedVisionLora::test_auto_dtype
618619
accuracy/test_llm_api_pytorch.py::TestPhi4MiniInstruct::test_auto_dtype
619620
accuracy/test_llm_api_pytorch.py::TestPhi4::test_auto_dtype
620621
accuracy/test_llm_api_pytorch.py::TestPhi4::test_fp8
@@ -675,14 +676,14 @@ test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistr
675676
test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-mixture_text_image-True]
676677
test_e2e.py::test_ptp_quickstart_multimodal[gemma-3-27b-it-gemma/gemma-3-27b-it-image-False]
677678
test_e2e.py::test_ptp_quickstart_multimodal[gemma-3-27b-it-gemma/gemma-3-27b-it-image-True]
678-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
679-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-0.8-image]
680-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-fp4-multimodals/Phi-4-multimodal-instruct-FP4-0.8-image]
681-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-fp8-multimodals/Phi-4-multimodal-instruct-FP8-0.8-image]
682-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
683-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-0.8-image]
684-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-fp8-multimodals/Phi-4-multimodal-instruct-FP8-0.8-image]
685-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-fp4-multimodals/Phi-4-multimodal-instruct-FP4-0.8-image]
679+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
680+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
681+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-fp4-multimodals/Phi-4-multimodal-instruct-FP4-image]
682+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-fp8-multimodals/Phi-4-multimodal-instruct-FP8-image]
683+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
684+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
685+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-fp8-multimodals/Phi-4-multimodal-instruct-FP8-image]
686+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-fp4-multimodals/Phi-4-multimodal-instruct-FP4-image]
686687
test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-audio]
687688
test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
688689
test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image_audio]

‎tests/integration/test_lists/qa/llm_function_l20.txt‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -43,6 +43,7 @@ accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype
4343
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype_long_rope
4444
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp4
4545
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp8
46+
accuracy/test_llm_api_pytorch.py::TestPhi4MMFusedVisionLora::test_auto_dtype
4647
accuracy/test_llm_api_pytorch.py::TestPhi4MiniInstruct::test_auto_dtype
4748
accuracy/test_llm_api_pytorch.py::TestMistralNemo12B::test_auto_dtype
4849

‎tests/integration/test_lists/qa/llm_function_nim.txt‎

Lines changed: 5 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -361,6 +361,7 @@ accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4[latency_moe_cu
361361
accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4[latency_moe_trtllm]
362362
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype
363363
accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_auto_dtype_long_rope
364+
accuracy/test_llm_api_pytorch.py::TestPhi4MMFusedVisionLora::test_auto_dtype
364365
accuracy/test_llm_api_pytorch.py::TestPhi4MiniInstruct::test_auto_dtype
365366
accuracy/test_llm_api_pytorch.py::TestPhi4::test_auto_dtype
366367
accuracy/test_llm_api_pytorch.py::TestPhi4::test_fp8
@@ -397,10 +398,10 @@ test_e2e.py::test_llmapi_generation_logits[llama-3.1-model/Llama-3.1-8B-Instruct
397398
test_e2e.py::test_llmapi_generation_logits[llama-3.1-model/Llama-3.1-8B-Instruct-False]
398399
test_e2e.py::test_llmapi_generation_logits[llama-3.3-models/Llama-3.3-70B-Instruct-True]
399400
test_e2e.py::test_llmapi_generation_logits[llama-3.3-models/Llama-3.3-70B-Instruct-False]
400-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
401-
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-0.8-image]
402-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
403-
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-0.8-image]
401+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
402+
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
403+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
404+
test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
404405

405406
examples/serve/test_serve.py::test_extra_llm_api_options
406407
examples/serve/test_serve_negative.py::test_invalid_max_tokens

‎tests/integration/test_lists/test-db/l0_h100.yml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -262,8 +262,8 @@ l0_h100:
262262
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_guided_decoding_with_ngram[llguidance]
263263
- test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image-True]
264264
- test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-mixture_text_image-True]
265-
- test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
266-
- test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-0.6-image]
265+
- test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
266+
- test_e2e.py::test_ptp_quickstart_multimodal_chunked_prefill[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image]
267267
- examples/test_mistral.py::test_mistral_with_bf16_lora_torch[mistral-7b-v0.1]
268268
- examples/test_phi.py::test_phi_4_mini_instruct_with_bf16_lora_torch[Phi-4-mini-instruct]
269269
- examples/test_llama.py::test_llama_3_x_with_bf16_lora_torch[llama-3.2-1b-instruct]

‎tests/integration/test_lists/waives.txt‎

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -359,8 +359,6 @@ accuracy/test_llm_api_pytorch_multimodal.py::TestLlava_V1_6_Mistral_7B::test_aut
359359
accuracy/test_disaggregated_serving.py::TestGPTOSS::test_auto_dtype[True] SKIP (https://nvbugs/5644632)
360360
accuracy/test_disaggregated_serving.py::TestGPTOSS::test_auto_dtype[False] SKIP (https://nvbugs/5644632)
361361
test_e2e.py::test_ptp_quickstart_multimodal_kv_cache_reuse[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-0.8-image] SKIP (https://nvbugs/5644190)
362-
test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image-True] SKIP (https://nvbugs/5648560)
363-
test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-image-False] SKIP (https://nvbugs/5648560)
364362
test_e2e.py::test_ptp_quickstart_multimodal_2gpu[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503] SKIP (https://nvbugs/5648560,https://nvbugs/5568836)
365363
accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[latency_trtllmgen_adp_lmtp] SKIP (https://nvbugs/5629136)
366364
perf/test_perf.py::test_perf[perf_sanity_upload-l0_dgx_b200] SKIP (https://nvbugs/5643646)
@@ -382,7 +380,6 @@ examples/test_multimodal.py::test_llm_multimodal_general[llava-onevision-qwen2-7
382380
examples/test_multimodal.py::test_llm_multimodal_general[llava-onevision-qwen2-7b-ov-hf-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:1] SKIP (https://nvbugs/5655832)
383381
examples/test_multimodal.py::test_llm_multimodal_general[Qwen2-VL-7B-Instruct-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:4] SKIP (https://nvbugs/5655832)
384382
disaggregated/test_disaggregated.py::test_disaggregated_mixed[TinyLlama-1.1B-Chat-v1.0] SKIP (https://nvbugs/5661926)
385-
test_e2e.py::test_ptp_quickstart_multimodal[mistral-small-3.1-24b-instruct-Mistral-Small-3.1-24B-Instruct-2503-mixture_text_image-True] SKIP (https://nvbugs/5568836)
386383
test_e2e.py::test_trtllm_multimodal_benchmark_serving SKIP (https://nvbugs/5647825)
387384
unittest/_torch/modules/test_fused_moe.py::test_fused_moe_alltoall_fp4[MNNVL] SKIP (https://nvbugs/5664904)
388385
unittest/_torch/modules/test_fused_moe.py::test_fused_moe_alltoall_fp4[DeepEP] SKIP (https://nvbugs/5664904)

0 commit comments

Comments
 (0)