Skip to content

[TRTLLM-6991][chore] add DeepSeek-R1 FP8 accuracy tests on Blackwell #6710

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 9 commits into from
Aug 19, 2025
Merged

[TRTLLM-6991][chore] add DeepSeek-R1 FP8 accuracy tests on Blackwell #6710

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
154c952
add DeepSeek-R1 FP8 accuracy tests on Blackwell.
lfr-0531 Aug 7, 2025
be5d6e1
add reference accuracy and moe_max_num_tokens.
lfr-0531 Aug 7, 2025
107baba
rm cuda graph padding.
lfr-0531 Aug 7, 2025
aa71210
update test lists.
lfr-0531 Aug 7, 2025
6277c69
update L0 test.
lfr-0531 Aug 11, 2025
83e9e47
fix num of tests in L0_Test.
lfr-0531 Aug 13, 2025
753b0b1
Merge branch 'main' into user/fanrongl/add_r1_fp8_blackwell_acc_test
litaotju Aug 14, 2025
97cf84a
Merge branch 'main' into user/fanrongl/add_r1_fp8_blackwell_acc_test
lfr-0531 Aug 16, 2025
12c0373
Merge branch 'main' into user/fanrongl/add_r1_fp8_blackwell_acc_test
lfr-0531 Aug 18, 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
13 changes: 7 additions & 6 deletions jenkins/L0_Test.groovy
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1892,12 +1892,13 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)

multiNodesSBSAConfigs = [
// Each stage test 1 testcase with 8 GPUs and 2 nodes.
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-1": ["gb200-multi-node", "l0_gb200_multi_nodes", 1, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-2": ["gb200-multi-node", "l0_gb200_multi_nodes", 2, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-3": ["gb200-multi-node", "l0_gb200_multi_nodes", 3, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-4": ["gb200-multi-node", "l0_gb200_multi_nodes", 4, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-5": ["gb200-multi-node", "l0_gb200_multi_nodes", 5, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-6": ["gb200-multi-node", "l0_gb200_multi_nodes", 6, 6, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-1": ["gb200-multi-node", "l0_gb200_multi_nodes", 1, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-2": ["gb200-multi-node", "l0_gb200_multi_nodes", 2, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-3": ["gb200-multi-node", "l0_gb200_multi_nodes", 3, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-4": ["gb200-multi-node", "l0_gb200_multi_nodes", 4, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-5": ["gb200-multi-node", "l0_gb200_multi_nodes", 5, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-6": ["gb200-multi-node", "l0_gb200_multi_nodes", 6, 7, 8, 2],
"GB200-8_GPUs-2_Nodes-PyTorch-Post-Merge-7": ["gb200-multi-node", "l0_gb200_multi_nodes", 7, 7, 8, 2],
]
fullSet += multiNodesSBSAConfigs.keySet()

Expand Down
3 changes: 3 additions & 0 deletions tests/integration/defs/accuracy/references/gsm8k.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,6 +74,9 @@ deepseek-ai/DeepSeek-R1:
- quant_algo: FP8_BLOCK_SCALES
spec_dec_algo: MTP
accuracy: 95.413
- quant_algo: FP8_BLOCK_SCALES
kv_cache_quant_algo: FP8
accuracy: 95.413
Qwen3/Qwen3-8B:
- accuracy: 87.1114
- quant_algo: FP8
Expand Down
3 changes: 3 additions & 0 deletions tests/integration/defs/accuracy/references/mmlu.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -163,6 +163,9 @@ deepseek-ai/DeepSeek-R1:
- quant_algo: FP8_BLOCK_SCALES
spec_dec_algo: MTP
accuracy: 87.573
- quant_algo: FP8_BLOCK_SCALES
kv_cache_quant_algo: FP8
accuracy: 87.573
Qwen3/Qwen3-8B:
- quant_algo: W4A8_MXFP4_FP8
accuracy: 72.70
Expand Down
9 changes: 8 additions & 1 deletion 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 @@ -1658,10 +1658,17 @@ def test_nvfp4_multi_gpus(self, tp_size, pp_size, ep_size, mtp_nextn, fp8kv,
def test_fp8_blockscale(self, tp_size, pp_size, ep_size, mtp_nextn, fp8kv,
attention_dp, cuda_graph, overlap_scheduler,
max_batch_size):
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.9)
if get_sm_version() == 100:
moe_config = MoeConfig(backend="DEEPGEMM", max_num_tokens=16384)
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.6)
else:
moe_config = MoeConfig()
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.9)

pytorch_config = dict(
disable_overlap_scheduler=not overlap_scheduler,
cuda_graph_config=CudaGraphConfig() if cuda_graph else None,
moe_config=moe_config,
)

if fp8kv:
Expand Down
1 change: 1 addition & 0 deletions tests/integration/test_lists/qa/llm_function_full.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -498,6 +498,7 @@ accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[latency_
accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[throughput]
accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[throughput_tp8]
accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[throughput_tp4]
accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_fp8_blockscale[throughput]
accuracy/test_llm_api_pytorch.py::TestQwen3_8B::test_fp8_block_scales[latency]
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_fp8_block_scales[latency-torch_compile=False]
accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_fp8_block_scales[latency-torch_compile=True]
Expand Down
1 change: 1 addition & 0 deletions tests/integration/test_lists/test-db/l0_gb200_multi_nodes.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@ l0_gb200_multi_nodes:
- accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[latency] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[throughput_tp8] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_nvfp4_multi_gpus[latency_trtllmgen] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestDeepSeekR1::test_fp8_blockscale[throughput] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4[latency_moe_cutlass] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4[latency_moe_trtllm] TIMEOUT (180)
- accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4[latency_moe_trtllm_eagle3] TIMEOUT (180)