mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
[None][infra] Prepare for single GPU GB200 test pipeline (#7073)
Signed-off-by: Yanchao Lu <yanchaol@nvidia.com>
This commit is contained in:
Yanchao Lu
GitHub
parent
48155f52bf
commit
ec35481b0a
@ -7,7 +7,6 @@ import groovy.json.JsonOutput
|
||||
import com.nvidia.bloom.KubernetesManager
|
||||
import com.nvidia.bloom.Constants
|
||||
import com.nvidia.bloom.CloudManager
|
||||
import com.nvidia.bloom.KubernetesManager
|
||||
import com.nvidia.bloom.SlurmConfig
|
||||
import com.nvidia.bloom.SlurmCluster
|
||||
import com.nvidia.bloom.SlurmPartition
|
||||
@ -219,8 +218,11 @@ def runLLMTestlistOnSlurm(pipeline, platform, testList, config=VANILLA_CONFIG, p
|
||||
SlurmPartition partition = SlurmConfig.partitionConfig[platform] as SlurmPartition
|
||||
SlurmCluster cluster = SlurmConfig.clusterConfig[partition.clusterName]
|
||||
|
||||
def nodeName = "${cluster.host}-test-${UUID.randomUUID().toString()}"
|
||||
def nodeSecret = CloudManager.createNode(nodeName)
|
||||
// Create a unique suffix for the node name and workspace
|
||||
String customSuffix = "${env.BUILD_TAG}-${UUID.randomUUID().toString().replaceAll("-", "").substring(0, 6)}".toLowerCase()
|
||||
def nodeName = "${cluster.host}-test-${customSuffix}"
|
||||
def customWorkspace = "/tmp/${nodeName}"
|
||||
def nodeSecret = CloudManager.createNode(nodeName, customWorkspace)
|
||||
|
||||
try {
|
||||
// Run ssh command to start node in desired cluster via SLURM
|
||||
@ -263,12 +265,30 @@ def runLLMTestlistOnSlurm(pipeline, platform, testList, config=VANILLA_CONFIG, p
|
||||
}
|
||||
|
||||
if (CloudManager.isNodeOnline(nodeName)) {
|
||||
def dockerArgs = "--gpus ${gpuCount} --cap-add=SYS_ADMIN --ipc=host --security-opt seccomp=unconfined -u root:root -v /home/scratch.trt_llm_data:/scratch.trt_llm_data:ro -v /tmp/ccache:${CCACHE_DIR}:rw -v /tmp/pipcache/http-v2:/root/.cache/pip/http-v2:rw --cap-add syslog"
|
||||
node(nodeName) {
|
||||
sh """
|
||||
env | sort
|
||||
pwd && ls -alh
|
||||
ls -alh ${env.WORKSPACE}
|
||||
ls -alh ${env.WORKSPACE_TMP}
|
||||
"""
|
||||
}
|
||||
|
||||
def dockerArgs = "--gpus ${gpuCount} " +
|
||||
"--cap-add=SYS_ADMIN " +
|
||||
"--ipc=host " +
|
||||
"--security-opt seccomp=unconfined " +
|
||||
"-u root:root " +
|
||||
"-v /home/scratch.trt_llm_data:/scratch.trt_llm_data:ro " +
|
||||
"-v /tmp/ccache:${CCACHE_DIR}:rw " +
|
||||
"-v /tmp/pipcache/http-v2:/root/.cache/pip/http-v2:rw " +
|
||||
"--cap-add syslog"
|
||||
|
||||
if (partition.clusterName == "dlcluster") {
|
||||
dockerArgs += " -e NVIDIA_IMEX_CHANNELS=0"
|
||||
}
|
||||
slurmRunner = runInDockerOnNodeMultiStage(LLM_DOCKER_IMAGE, nodeName, dockerArgs, false)
|
||||
|
||||
slurmRunner = runInDockerOnNodeMultiStage(LLM_DOCKER_IMAGE, nodeName, dockerArgs, true)
|
||||
executeLLMTestOnSlurm(pipeline, platform, testList, config, perfMode, stageName, splitId, splits, skipInstallWheel, cpver, slurmRunner)
|
||||
} else {
|
||||
echo "The node does not come online in 2 hours, terminating the job"
|
||||
@ -560,6 +580,13 @@ def cacheErrorAndUploadResult(stageName, taskRunner, finallyRunner, noResultIfSu
|
||||
"${UPLOAD_PATH}/test-results/"
|
||||
)
|
||||
junit(testResults: "${stageName}/results*.xml")
|
||||
|
||||
// Clean up the workspace
|
||||
sh """
|
||||
env | sort
|
||||
pwd && ls -alh
|
||||
rm -rf ./*
|
||||
"""
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -796,7 +823,7 @@ def echoNodeAndGpuInfo(pipeline, stageName)
|
||||
|
||||
def runLLMDocBuild(pipeline, config)
|
||||
{
|
||||
// Step 1: cloning tekit source code
|
||||
// Step 1: cloning source code
|
||||
sh "pwd && ls -alh"
|
||||
sh "env | sort"
|
||||
// allow to checkout from forked repo, svc_tensorrt needs to have access to the repo, otherwise clone will fail
|
||||
@ -1241,13 +1268,16 @@ def rerunFailedTests(stageName, llmSrc, testCmdLine) {
|
||||
|
||||
def runLLMTestlistOnPlatformImpl(pipeline, platform, testList, config=VANILLA_CONFIG, perfMode=false, stageName="Undefined", splitId=1, splits=1, skipInstallWheel=false, cpver="cp312")
|
||||
{
|
||||
// Step 1: create LLM_ROOT dir
|
||||
sh "pwd && ls -alh"
|
||||
// TODO: proper way to clean workspace, maybe save in a folder named with BUILD_ID.
|
||||
// So that it can work with multiple job running in same node
|
||||
sh "rm -rf ./*"
|
||||
// Step 1: create LLM_ROOT dir and clean up the workspace
|
||||
def llmRootConfig = "${LLM_ROOT}${config}"
|
||||
sh "mkdir ${llmRootConfig}"
|
||||
sh """
|
||||
env | sort
|
||||
pwd && ls -alh
|
||||
rm -rf ./*
|
||||
mkdir ${llmRootConfig}
|
||||
ls -alh ${env.WORKSPACE}
|
||||
ls -alh ${env.WORKSPACE_TMP}
|
||||
"""
|
||||
|
||||
def llmPath = sh (script: "realpath ${llmRootConfig}", returnStdout: true).trim()
|
||||
def llmSrc = "${llmPath}/TensorRT-LLM/src"
|
||||
@ -1765,7 +1795,6 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)
|
||||
"DGX_H100-4_GPUs-PyTorch-DeepSeek-1": ["dgx-h100-x4", "l0_dgx_h100", 1, 2, 4],
|
||||
"DGX_H100-4_GPUs-PyTorch-DeepSeek-2": ["dgx-h100-x4", "l0_dgx_h100", 2, 2, 4],
|
||||
"DGX_H100-4_GPUs-PyTorch-Others-1": ["dgx-h100-x4", "l0_dgx_h100", 1, 1, 4],
|
||||
"DGX_H100-4_GPUs-Triton-Post-Merge-1": ["dgx-h100-x4", "l0_dgx_h100", 1, 1, 4],
|
||||
"DGX_H100-4_GPUs-CPP-1": ["dgx-h100-x4", "l0_dgx_h100", 1, 1, 4],
|
||||
"A10-PyTorch-1": ["a10", "l0_a10", 1, 1],
|
||||
"A10-CPP-1": ["a10", "l0_a10", 1, 1],
|
||||
@ -1838,6 +1867,7 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)
|
||||
"B200_PCIe-TensorRT-Post-Merge-2": ["b100-ts2", "l0_b200", 2, 2],
|
||||
"H100_PCIe-TensorRT-Perf-1": ["h100-cr", "l0_perf", 1, 1],
|
||||
"H100_PCIe-PyTorch-Perf-1": ["h100-cr", "l0_perf", 1, 1],
|
||||
"DGX_H200-4_GPUs-Triton-Post-Merge-1": ["dgx-h200-x4", "l0_dgx_h200", 1, 1, 4],
|
||||
"DGX_H200-8_GPUs-PyTorch-Post-Merge-1": ["dgx-h200-x8", "l0_dgx_h200", 1, 1, 8],
|
||||
"DGX_H200-4_GPUs-PyTorch-Post-Merge-1": ["dgx-h200-x4", "l0_dgx_h200", 1, 1, 4],
|
||||
"DGX_H200-4_GPUs-TensorRT-Post-Merge-1": ["dgx-h200-x4", "l0_dgx_h200", 1, 3, 4],
|
||||
@ -1890,8 +1920,10 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)
|
||||
fullSet += SBSATestConfigs.keySet()
|
||||
|
||||
SBSASlurmTestConfigs = [
|
||||
"GB200-4_GPUs-PyTorch-1": ["gb200-x4", "l0_gb200", 1, 1, 4],
|
||||
"GB200-4_GPUs-PyTorch-Post-Merge-1": ["gb200-x4", "l0_gb200", 1, 1, 4],
|
||||
// Not used in the pipeline now
|
||||
// "GB200-PyTorch-1": ["gb200-single", "l0_gb200", 1, 3],
|
||||
"GB200-4_GPUs-PyTorch-1": ["gb200-x4", "l0_gb200_multi_gpus", 1, 1, 4],
|
||||
"GB200-4_GPUs-PyTorch-Post-Merge-1": ["gb200-x4", "l0_gb200_multi_gpus", 1, 1, 4],
|
||||
]
|
||||
fullSet += SBSASlurmTestConfigs.keySet()
|
||||
|
||||
@ -1903,7 +1935,6 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)
|
||||
"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()
|
||||
|
||||
@ -2123,7 +2154,9 @@ def launchTestJobs(pipeline, testFilter, dockerNode=null)
|
||||
echo "###### Check pip install Start ######"
|
||||
withEnv(libEnv) {
|
||||
sh "env | sort"
|
||||
checkPipInstall(pipeline, "${cpu_arch}/${wheelPath}")
|
||||
timeout(time: 1, unit: 'HOURS') {
|
||||
checkPipInstall(pipeline, "${cpu_arch}/${wheelPath}")
|
||||
}
|
||||
}
|
||||
echo "###### Run LLMAPI tests Start ######"
|
||||
def config = VANILLA_CONFIG
|
||||
@ -2458,7 +2491,7 @@ pipeline {
|
||||
|
||||
def testPhase2StageName = env.testPhase2StageName
|
||||
if (testPhase2StageName) {
|
||||
def dgxSigns = ["DGX_H100", "DGX_H200", "GB200", "DGX_B200", "RTXPro6000-4_GPUs"]
|
||||
def dgxSigns = ["2_GPUs", "4_GPUs", "8_GPUs"]
|
||||
singleGpuJobs = parallelJobs.findAll{!dgxSigns.any{sign -> it.key.contains(sign)}}
|
||||
dgxJobs = parallelJobs.findAll{dgxSigns.any{sign -> it.key.contains(sign)}}
|
||||
}
|
||||
|
||||
@ -34,7 +34,7 @@ else
|
||||
done
|
||||
fi
|
||||
testList="$testList_$splitId"
|
||||
export CPP_TEST_TIMEOUT_OVERRIDDEN=7200
|
||||
export CPP_TEST_TIMEOUT_OVERRIDDEN=$pytestTestTimeout
|
||||
export LLM_ROOT=$llmSrcNode
|
||||
export LLM_MODELS_ROOT=$MODEL_CACHE_DIR
|
||||
export UCX_TLS=^gdr_copy
|
||||
@ -43,6 +43,7 @@ testCmdLines=(
|
||||
"$llmSrcNode/tensorrt_llm/llmapi/trtllm-llmapi-launch"
|
||||
"pytest"
|
||||
"-v"
|
||||
"--timeout-method=thread"
|
||||
"--timeout=$pytestTestTimeout"
|
||||
"--test-list=$testListPathNode"
|
||||
"--waives-file=$waivesListPathNode"
|
||||
|
||||
@ -209,18 +209,3 @@ l0_dgx_h100:
|
||||
- cpp/test_multi_gpu.py::TestDisagg::test_spawn_orchestrator[llama-ucx_kvcache-90]
|
||||
- cpp/test_multi_gpu.py::TestDisagg::test_orchestrator_params[llama-nixl_kvcache-90] TIMEOUT (90)
|
||||
- cpp/test_multi_gpu.py::TestDisagg::test_spawn_orchestrator[llama-nixl_kvcache-90]
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*h100*'
|
||||
linux_distribution_name: ubuntu*
|
||||
terms:
|
||||
stage: post_merge
|
||||
backend: triton
|
||||
auto_trigger: others
|
||||
tests:
|
||||
- triton_server/test_triton_llm.py::test_llmapi_backend[4-0-disableDecoupleMode-tensorrt_llm]
|
||||
|
||||
@ -166,3 +166,19 @@ l0_dgx_h200:
|
||||
- examples/test_enc_dec.py::test_llm_enc_dec_general[compare_hf-t5-small-float16-enable_gemm_plugin-enable_attention_plugin-disable_paged_kv_cache-tp:2-pp:2-nb:1-disable_fp8]
|
||||
- examples/test_gpt.py::test_llm_gpt2_next_prompt_tuning[use_py_session-tp2]
|
||||
- unittest/llmapi/apps/_test_openai_multi_gpu.py -m "part0"
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*h200*'
|
||||
linux_distribution_name: ubuntu*
|
||||
cpu: x86_64
|
||||
terms:
|
||||
stage: post_merge
|
||||
backend: triton
|
||||
tests:
|
||||
# ------------- Triton tests ---------------
|
||||
- triton_server/test_triton_llm.py::test_llmapi_backend[4-0-disableDecoupleMode-tensorrt_llm]
|
||||
|
||||
@ -1,10 +1,12 @@
|
||||
# Don't add any tests here.
|
||||
# Copied from l0_b200.yml but not used in the pipeline now
|
||||
version: 0.0.1
|
||||
l0_gb200:
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
gte: 1
|
||||
lte: 1
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
@ -14,33 +16,111 @@ l0_gb200:
|
||||
stage: pre_merge
|
||||
backend: pytorch
|
||||
tests:
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[tp4-attn_backend=TRTLLM-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[tp4-attn_backend=FLASHINFER-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[pp4-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=False-attn_backend=TRTLLM-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=False-attn_backend=FLASHINFER-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[pp4-fp8kv=True-attn_backend=TRTLLM-torch_compile=False]
|
||||
# ------------- PyTorch tests ---------------
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4_streaming[stream_interval_4]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4_streaming[stream_interval_64]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16[mtp_nextn=0-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False-enable_chunked_prefill=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16[mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False-enable_chunked_prefill=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16[mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False-enable_chunked_prefill=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[tp4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[ep4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=0-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=True-attention_dp=False-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=TRTLLM-mtp_nextn=0-fp8kv=True-attention_dp=False-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=2-fp8kv=True-attention_dp=False-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=TRTLLM-mtp_nextn=2-fp8kv=True-attention_dp=False-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_no_kv_cache_reuse[quant_dtype=none-mtp_nextn=2-fp8kv=False-attention_dp=True-cuda_graph=True-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_no_kv_cache_reuse[quant_dtype=nvfp4-mtp_nextn=0-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_chunked_prefill[quant_dtype=none-kv_cache_reuse=True-fp8kv=False-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_chunked_prefill[quant_dtype=none-kv_cache_reuse=False-fp8kv=False-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_chunked_prefill[quant_dtype=nvfp4-kv_cache_reuse=True-fp8kv=False-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_chunked_prefill[quant_dtype=nvfp4-kv_cache_reuse=True-fp8kv=True-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_8B::test_w4a8_mxfp4[fp8-latency]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_8B::test_w4a8_mxfp4[mxfp8-latency]
|
||||
- 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]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[fp8-latency-CUTLASS]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[fp8-latency-TRITON]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[fp8-latency-TRTLLM]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[mxfp8-latency-TRTLLM]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[mxfp8-latency-CUTLASS]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a16_mxfp4[latency-TRTLLM]
|
||||
- accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[True-True-cutlass]
|
||||
- accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[True-True-trtllm]
|
||||
- accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[True-True-triton]
|
||||
- disaggregated/test_workers.py::test_workers_kv_cache_aware_router_eviction[TinyLlama-1.1B-Chat-v1.0] # nvbugs 5300551
|
||||
- test_e2e.py::test_ptp_quickstart_advanced[Llama3.1-8B-NVFP4-nvfp4-quantized/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]
|
||||
- test_e2e.py::test_ptp_quickstart_advanced_mtp[DeepSeek-V3-Lite-BF16-DeepSeek-V3-Lite/bf16]
|
||||
- test_e2e.py::test_ptp_quickstart_advanced_mixed_precision
|
||||
- test_e2e.py::test_ptp_quickstart_advanced_eagle3[Llama-3.1-8b-Instruct-llama-3.1-model/Llama-3.1-8B-Instruct-EAGLE3-LLaMA3.1-Instruct-8B]
|
||||
- test_e2e.py::test_ptp_quickstart_advanced_ngram[Llama-3.1-8B-Instruct-llama-3.1-model/Llama-3.1-8B-Instruct]
|
||||
- test_e2e.py::test_trtllm_bench_pytorch_backend_sanity[meta-llama/Llama-3.1-8B-llama-3.1-8b-False-False]
|
||||
- unittest/_torch/attention
|
||||
- unittest/_torch/compilation
|
||||
- unittest/_torch/debugger
|
||||
- unittest/_torch/executor
|
||||
- unittest/_torch/misc
|
||||
- unittest/_torch/modules
|
||||
- unittest/_torch/multimodal
|
||||
- unittest/_torch/sampler
|
||||
- unittest/_torch/speculative
|
||||
- unittest/_torch/thop
|
||||
- unittest/_torch/modeling -k "modeling_llama"
|
||||
- unittest/_torch/modeling -k "modeling_mixtral"
|
||||
- unittest/_torch/modeling -k "modeling_deepseek"
|
||||
- unittest/_torch/modeling -k "modeling_gpt_oss"
|
||||
- unittest/_torch/auto_deploy/unit/singlegpu -k "not test_trtllm_bench_backend_comparison"
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
gte: 1
|
||||
lte: 1
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
linux_distribution_name: ubuntu*
|
||||
cpu: aarch64
|
||||
terms:
|
||||
stage: post_merge
|
||||
backend: tensorrt
|
||||
tests:
|
||||
# ------------- TRT tests ---------------
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_nvfp4
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_nvfp4_gemm_plugin[disable_norm_quant_fusion-disable_fused_quant]
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_nvfp4_gemm_plugin[disable_norm_quant_fusion-enable_fused_quant]
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_nvfp4_gemm_plugin[enable_norm_quant_fusion-disable_fused_quant]
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_nvfp4_gemm_plugin[enable_norm_quant_fusion-enable_fused_quant]
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_auto_dtype
|
||||
- accuracy/test_cli_flow.py::TestLlama3_8BInstruct::test_fp8
|
||||
- accuracy/test_cli_flow.py::TestMixtral8x7B::test_nvfp4_prequantized
|
||||
- unittest/trt/attention/test_gpt_attention.py -k "trtllm_gen"
|
||||
- unittest/llmapi/test_llm_quant.py
|
||||
- unittest/trt/functional/test_fp4_gemm.py
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 1
|
||||
lte: 1
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
linux_distribution_name: ubuntu*
|
||||
cpu: aarch64
|
||||
terms:
|
||||
stage: post_merge
|
||||
backend: triton
|
||||
tests:
|
||||
# ------------- Triton tests ---------------
|
||||
- triton_server/test_triton.py::test_llava[llava]
|
||||
- triton_server/test_triton.py::test_gpt_ib_ptuning[gpt-ib-ptuning]
|
||||
- triton_server/test_triton.py::test_gpt_2b_ib_lora[gpt-2b-ib-lora]
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 1
|
||||
lte: 1
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
@ -50,23 +130,16 @@ l0_gb200:
|
||||
stage: post_merge
|
||||
backend: pytorch
|
||||
tests:
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=True-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=True-attn_backend=FLASHINFER-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[pp4-fp8kv=False-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=2-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=2-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus_online_eplb[fp8kv=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus_online_eplb[mtp_nextn=2]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4_4gpus[latency_moe_trtllm_eagle3] TIMEOUT (90)
|
||||
# ------------- PyTorch tests ---------------
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=True-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=2-fp8kv=True-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=True-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=True-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=0-fp8kv=False-attention_dp=False-cuda_graph=True-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=TRTLLM-mtp_nextn=0-fp8kv=False-attention_dp=False-cuda_graph=True-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=CUTLASS-mtp_nextn=2-fp8kv=False-attention_dp=False-cuda_graph=True-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4[moe_backend=TRTLLM-mtp_nextn=2-fp8kv=False-attention_dp=False-cuda_graph=True-overlap_scheduler=False-torch_compile=False]
|
||||
|
||||
69
tests/integration/test_lists/test-db/l0_gb200_multi_gpus.yml
Normal file
69
tests/integration/test_lists/test-db/l0_gb200_multi_gpus.yml
Normal file
@ -0,0 +1,69 @@
|
||||
version: 0.0.1
|
||||
l0_gb200_multi_gpus:
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
linux_distribution_name: ubuntu*
|
||||
cpu: aarch64
|
||||
terms:
|
||||
stage: pre_merge
|
||||
backend: pytorch
|
||||
tests:
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[tp4-attn_backend=TRTLLM-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[tp4-attn_backend=FLASHINFER-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16_4gpus[pp4-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=False-attn_backend=TRTLLM-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=False-attn_backend=FLASHINFER-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[pp4-fp8kv=True-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[tp4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[ep4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=0-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- condition:
|
||||
ranges:
|
||||
system_gpu_count:
|
||||
gte: 4
|
||||
lte: 4
|
||||
wildcards:
|
||||
gpu:
|
||||
- '*gb200*'
|
||||
linux_distribution_name: ubuntu*
|
||||
cpu: aarch64
|
||||
terms:
|
||||
stage: post_merge
|
||||
backend: pytorch
|
||||
tests:
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=True-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[tp4-fp8kv=True-attn_backend=FLASHINFER-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_fp8_4gpus[pp4-fp8kv=False-attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-ep4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp2pp2-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-tp2pp2-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-pp4-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=2-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=2-tp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=TRTLLM-mtp_nextn=2-ep4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus_online_eplb[fp8kv=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus_online_eplb[mtp_nextn=2]
|
||||
- accuracy/test_llm_api_pytorch.py::TestQwen3_235B_A22B::test_nvfp4_4gpus[latency_moe_trtllm_eagle3] TIMEOUT (90)
|
||||
Loading…
Reference in New Issue
Block a user