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 15, 19:45 UTC
Re-enable the AMD Entrypoints Test (#14711)
Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
Alexei-V-Ivanov-AMD committed
Mar 14, 2025 at 14:18 UTC
270a5da495d24e947a71e2fa0c56635f4fad2dc3
3 files changed
+43
-5
.buildkite/run-amd-test.sh
+19
-5
@@ -101,16 +101,30 @@ if [[ $commands == *" kernels "* ]]; then
101
--ignore=kernels/test_permute_cols.py"
102
fi
103
104
-#ignore certain Entrypoints tests
104
+#ignore certain Entrypoints/openai tests
105
if [[ $commands == *" entrypoints/openai "* ]]; then
106
commands=${commands//" entrypoints/openai "/" entrypoints/openai \
107
- --ignore=entrypoints/openai/test_accuracy.py \
107
--ignore=entrypoints/openai/test_audio.py \
109
- --ignore=entrypoints/openai/test_encoder_decoder.py \
110
- --ignore=entrypoints/openai/test_embedding.py \
111
- --ignore=entrypoints/openai/test_oot_registration.py "}
108
+ --ignore=entrypoints/openai/test_chat.py \
109
+ --ignore=entrypoints/openai/test_shutdown.py \
110
+ --ignore=entrypoints/openai/test_completion.py \
111
+ --ignore=entrypoints/openai/test_sleep.py \
112
+ --ignore=entrypoints/openai/test_models.py \
113
+ --ignore=entrypoints/openai/test_prompt_validation.py "}
114
fi
115
116
+#ignore certain Entrypoints/llm tests
117
+if [[ $commands == *" && pytest -v -s entrypoints/llm/test_guided_generate.py"* ]]; then
118
+ commands=${commands//" && pytest -v -s entrypoints/llm/test_guided_generate.py"/" "}
119
+fi
120
+
121
+# --ignore=entrypoints/openai/test_encoder_decoder.py \
122
+# --ignore=entrypoints/openai/test_embedding.py \
123
+# --ignore=entrypoints/openai/test_oot_registration.py
124
+# --ignore=entrypoints/openai/test_accuracy.py \
125
+# --ignore=entrypoints/openai/test_models.py <= Fails on MI250 but passes on MI300 as of 2025-03-13
126
+
127
+
128
PARALLEL_JOB_COUNT=8
129
# check if the command contains shard flag, we will run all shards in parallel because the host have 8 GPUs.
130
if [[ $commands == *"--shard-id="* ]]; then
Dockerfile.rocm
+1
@@ -61,6 +61,7 @@ RUN python3 -m pip install --upgrade pip && rm -rf /var/lib/apt/lists/*
61
RUN --mount=type=bind,from=export_vllm,src=/,target=/install \
62
cd /install \
63
&& pip install -U -r requirements/rocm.txt \
64
+ && pip install -U -r requirements/rocm-test.txt \
65
&& pip uninstall -y vllm \
66
&& pip install *.whl
67
requirements/rocm-test.txt
new
+23
@@ -0,0 +1,23 @@
1
+
2
+# entrypoints test
3
+# librosa==0.10.2.post1 # required by audio tests in entrypoints/openai
4
+audioread==3.0.1
5
+cffi==1.17.1
6
+decorator==5.2.1
7
+lazy-loader==0.4
8
+platformdirs==4.3.6
9
+pooch==1.8.2
10
+#pycparse==2.22
11
+soundfile==0.13.1
12
+soxr==0.5.0.post1
13
+librosa==0.10.2.post1
14
+
15
+# entrypoints test
16
+#vllm[video] # required by entrypoints/openai/test_video.py
17
+decord==0.6.0
18
+
19
+# entrypoints test
20
+#sentence-transformers # required by entrypoints/openai/test_score.py
21
+sentence-transformers==3.4.1
22
+
23
+