A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 20:15 UTC
[CI][NIXL] Split DPEP tests (#31491)
Signed-off-by: NickLucche <nlucches@redhat.com>
Nicolò Lucchesi committed
Dec 30, 2025 at 13:26 UTC
ab1af6aa3e8cac44725b6502f8afd107de18cec4
2 files changed
+25
-4
.buildkite/test-pipeline.yaml
+12
-1
@@ -1276,7 +1276,18 @@ steps:
1276
- tests/v1/kv_connector/nixl_integration/
1277
commands:
1278
- uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
1279
- - bash v1/kv_connector/nixl_integration/tp_config_sweep_accuracy_test.sh
1279
+ - bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
1280
+
1281
+- label: NixlConnector PD accuracy tests (Distributed) # TODO
1282
+ timeout_in_minutes: 30
1283
+ working_dir: "/vllm-workspace/tests"
1284
+ num_gpus: 4
1285
+ source_file_dependencies:
1286
+ - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
1287
+ - tests/v1/kv_connector/nixl_integration/
1288
+ commands:
1289
+ - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
1290
+ - DP_EP=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
1291
1292
1293
##### multi gpus test #####
tests/v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
renamed
+13
-3
@@ -5,16 +5,26 @@ set -euo pipefail
5
SCRIPT="v1/kv_connector/nixl_integration/run_accuracy_test.sh"
6
7
# Define test configurations
8
-configs=(
8
+tp_configs=(
9
"GPU_MEMORY_UTILIZATION=0.6 PREFILLER_TP_SIZE=2 DECODER_TP_SIZE=2"
10
"GPU_MEMORY_UTILIZATION=0.6 PREFILLER_TP_SIZE=1 DECODER_TP_SIZE=2"
11
"GPU_MEMORY_UTILIZATION=0.6 PREFILLER_TP_SIZE=2 DECODER_TP_SIZE=1"
12
"GPU_MEMORY_UTILIZATION=0.8 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny" # MLA case
13
"GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=1 DECODER_TP_SIZE=2 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny"
14
"GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=2 DECODER_TP_SIZE=1 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny"
15
- "DP_EP=1 GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=1 DECODER_TP_SIZE=2 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny" # MLA+P-TP1, D-DPEP=2 (TP=1)
16
- "DP_EP=1 GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=2 DECODER_TP_SIZE=2 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny" # MLA+P-TP2, D-DPEP=2 (TP=1)
15
)
16
+dp_ep_configs=(
17
+"DP_EP=1 GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=1 DECODER_TP_SIZE=2 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny" # MLA+P-TP1, D-DPEP=2 (TP=1)
18
+"DP_EP=1 GPU_MEMORY_UTILIZATION=0.8 PREFILLER_TP_SIZE=2 DECODER_TP_SIZE=2 MODEL_NAMES=deepseek-ai/deepseek-vl2-tiny" # MLA+P-TP2, D-DPEP=2 (TP=1)
19
+)
20
+
21
+# Select config array based on DP_EP env var
22
+if [[ -n "${DP_EP:-}" ]]; then
23
+ configs=("${dp_ep_configs[@]}")
24
+ echo "DP_EP is set, using dp_ep_configs"
25
+else
26
+ configs=("${tp_configs[@]}")
27
+fi
28
29
run_tests() {
30
local label=$1