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, 16:00 UTC
Re-enable the AMD Testing for the passing tests. (#15586)
Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
Alexei-V-Ivanov-AMD committed
Apr 3, 2025 at 13:05 UTC
03a70eacafccd0ed4688ed77dc86e2551ffa1c04
2 files changed
+26
-9
.buildkite/run-amd-test.sh
+17
-3
@@ -105,19 +105,33 @@ fi
105
if [[ $commands == *" entrypoints/openai "* ]]; then
106
commands=${commands//" entrypoints/openai "/" entrypoints/openai \
107
--ignore=entrypoints/openai/test_audio.py \
108
- --ignore=entrypoints/openai/test_chat.py \
108
--ignore=entrypoints/openai/test_shutdown.py \
109
--ignore=entrypoints/openai/test_completion.py \
110
--ignore=entrypoints/openai/test_sleep.py \
111
--ignore=entrypoints/openai/test_models.py \
112
+ --ignore=entrypoints/openai/test_lora_adapters.py \
113
+ --ignore=entrypoints/openai/test_return_tokens_as_ids.py \
114
+ --ignore=entrypoints/openai/test_root_path.py \
115
+ --ignore=entrypoints/openai/test_tokenization.py \
116
--ignore=entrypoints/openai/test_prompt_validation.py "}
117
fi
118
119
#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"/" "}
120
+if [[ $commands == *" entrypoints/llm "* ]]; then
121
+ commands=${commands//" entrypoints/llm "/" entrypoints/llm \
122
+ --ignore=entrypoints/llm/test_chat.py \
123
+ --ignore=entrypoints/llm/test_accuracy.py \
124
+ --ignore=entrypoints/llm/test_init.py \
125
+ --ignore=entrypoints/llm/test_generate_multiple_loras.py \
126
+ --ignore=entrypoints/llm/test_prompt_validation.py "}
127
fi
128
129
+#Obsolete currently
130
+##ignore certain Entrypoints/llm tests
131
+#if [[ $commands == *" && pytest -v -s entrypoints/llm/test_guided_generate.py"* ]]; then
132
+# commands=${commands//" && pytest -v -s entrypoints/llm/test_guided_generate.py"/" "}
133
+#fi
134
+
135
# --ignore=entrypoints/openai/test_encoder_decoder.py \
136
# --ignore=entrypoints/openai/test_embedding.py \
137
# --ignore=entrypoints/openai/test_oot_registration.py
.buildkite/test-pipeline.yaml
+9
-6
@@ -104,7 +104,7 @@ steps:
104
- label: Entrypoints Test # 40min
105
working_dir: "/vllm-workspace/tests"
106
fast_check: true
107
- mirror_hardwares: [amd]
107
+ #mirror_hardwares: [amd]
108
source_file_dependencies:
109
- vllm/
110
- tests/entrypoints/llm
@@ -155,6 +155,7 @@ steps:
155
- popd
156
157
- label: Metrics, Tracing Test # 10min
158
+ mirror_hardwares: [amd]
159
num_gpus: 2
160
source_file_dependencies:
161
- vllm/
@@ -173,7 +174,7 @@ steps:
174
##### 1 GPU test #####
175
176
- label: Regression Test # 5min
176
- mirror_hardwares: [amd]
177
+ #mirror_hardwares: [amd]
178
source_file_dependencies:
179
- vllm/
180
- tests/test_regression
@@ -284,7 +285,7 @@ steps:
285
- pytest -v -s spec_decode/e2e/test_eagle_correctness.py
286
287
- label: LoRA Test %N # 15min each
287
- mirror_hardwares: [amd]
288
+ #mirror_hardwares: [amd]
289
source_file_dependencies:
290
- vllm/lora
291
- tests/lora
@@ -310,7 +311,7 @@ steps:
311
- pytest -v -s compile/test_full_graph.py
312
313
- label: Kernels Test %N # 1h each
313
- mirror_hardwares: [amd]
314
+ # mirror_hardwares: [amd]
315
source_file_dependencies:
316
- csrc/
317
- vllm/attention
@@ -320,7 +321,7 @@ steps:
321
parallelism: 4
322
323
- label: Tensorizer Test # 11min
323
- mirror_hardwares: [amd]
324
+ # mirror_hardwares: [amd]
325
soft_fail: true
326
source_file_dependencies:
327
- vllm/model_executor/model_loader
@@ -371,7 +372,7 @@ steps:
372
373
- label: OpenAI-Compatible Tool Use # 20 min
374
fast_check: false
374
- mirror_hardwares: [ amd ]
375
+ #mirror_hardwares: [ amd ]
376
source_file_dependencies:
377
- vllm/
378
- tests/tool_use
@@ -463,6 +464,7 @@ steps:
464
465
# This test is used only in PR development phase to test individual models and should never run on main
466
- label: Custom Models Test
467
+ mirror_hardwares: [amd]
468
optional: true
469
commands:
470
- echo 'Testing custom models...'
@@ -474,6 +476,7 @@ steps:
476
##### multi gpus test #####
477
478
- label: Distributed Comm Ops Test # 7min
479
+ mirror_hardwares: [amd]
480
working_dir: "/vllm-workspace/tests"
481
num_gpus: 2
482
source_file_dependencies: