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, 12:15 UTC
Enabling multi-group kernel tests. (#17115)
Enabling multi-group kernel tests. (#17115) Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
Alexei-V-Ivanov-AMD committed
Apr 29, 2025 at 12:27 UTC
608968b7c5709dc74754cb09c72a5793dec7af2b
2 files changed
+49
-30
.buildkite/scripts/hardware_ci/run-amd-test.sh
+44
-30
@@ -75,37 +75,51 @@ HF_MOUNT="/root/.cache/huggingface"
75
commands=$@
76
echo "Commands:$commands"
77
#ignore certain kernels tests
78
-if [[ $commands == *" kernels "* ]]; then
78
+if [[ $commands == *" kernels/core"* ]]; then
79
commands="${commands} \
80
- --ignore=kernels/test_attention_selector.py \
81
- --ignore=kernels/test_blocksparse_attention.py \
82
- --ignore=kernels/test_causal_conv1d.py \
83
- --ignore=kernels/test_cutlass.py \
84
- --ignore=kernels/test_encoder_decoder_attn.py \
85
- --ignore=kernels/test_flash_attn.py \
86
- --ignore=kernels/test_flashinfer.py \
87
- --ignore=kernels/test_int8_quant.py \
88
- --ignore=kernels/test_machete_gemm.py \
89
- --ignore=kernels/test_mamba_ssm.py \
90
- --ignore=kernels/test_marlin_gemm.py \
91
- --ignore=kernels/test_moe.py \
92
- --ignore=kernels/test_prefix_prefill.py \
93
- --ignore=kernels/test_rand.py \
94
- --ignore=kernels/test_sampler.py \
95
- --ignore=kernels/test_cascade_flash_attn.py \
96
- --ignore=kernels/test_mamba_mixer2.py \
97
- --ignore=kernels/test_aqlm.py \
98
- --ignore=kernels/test_machete_mm.py \
99
- --ignore=kernels/test_mha_attn.py \
100
- --ignore=kernels/test_block_fp8.py \
101
- --ignore=kernels/test_cutlass_moe.py \
102
- --ignore=kernels/test_mamba_ssm_ssd.py \
103
- --ignore=kernels/test_attention.py \
104
- --ignore=kernels/test_block_int8.py \
105
- --ignore=kernels/test_fused_quant_layernorm.py \
106
- --ignore=kernels/test_int8_kernel.py \
107
- --ignore=kernels/test_triton_moe_ptpc_fp8.py \
108
- --ignore=kernels/test_permute_cols.py"
80
+ --ignore=kernels/core/test_fused_quant_layernorm.py \
81
+ --ignore=kernels/core/test_permute_cols.py"
82
+fi
83
+
84
+if [[ $commands == *" kernels/attention"* ]]; then
85
+ commands="${commands} \
86
+ --ignore=kernels/attention/stest_attention_selector.py \
87
+ --ignore=kernels/attention/test_blocksparse_attention.py \
88
+ --ignore=kernels/attention/test_encoder_decoder_attn.py \
89
+ --ignore=kernels/attention/test_attention_selector.py \
90
+ --ignore=kernels/attention/test_flash_attn.py \
91
+ --ignore=kernels/attention/test_flashinfer.py \
92
+ --ignore=kernels/attention/test_prefix_prefill.py \
93
+ --ignore=kernels/attention/test_cascade_flash_attn.py \
94
+ --ignore=kernels/attention/test_mha_attn.py \
95
+ --ignore=kernels/attention/test_lightning_attn.py \
96
+ --ignore=kernels/attention/test_attention.py"
97
+fi
98
+
99
+if [[ $commands == *" kernels/quantization"* ]]; then
100
+ commands="${commands} \
101
+ --ignore=kernels/quantization/test_int8_quant.py \
102
+ --ignore=kernels/quantization/test_aqlm.py \
103
+ --ignore=kernels/quantization/test_machete_mm.py \
104
+ --ignore=kernels/quantization/test_block_fp8.py \
105
+ --ignore=kernels/quantization/test_block_int8.py \
106
+ --ignore=kernels/quantization/test_marlin_gemm.py \
107
+ --ignore=kernels/quantization/test_cutlass_scaled_mm.py \
108
+ --ignore=kernels/quantization/test_int8_kernel.py"
109
+fi
110
+
111
+if [[ $commands == *" kernels/mamba"* ]]; then
112
+ commands="${commands} \
113
+ --ignore=kernels/mamba/test_mamba_mixer2.py \
114
+ --ignore=kernels/mamba/test_causal_conv1d.py \
115
+ --ignore=kernels/mamba/test_mamba_ssm_ssd.py"
116
+fi
117
+
118
+if [[ $commands == *" kernels/moe"* ]]; then
119
+ commands="${commands} \
120
+ --ignore=kernels/moe/test_moe.py \
121
+ --ignore=kernels/moe/test_cutlass_moe.py \
122
+ --ignore=kernels/moe/test_triton_moe_ptpc_fp8.py"
123
fi
124
125
#ignore certain Entrypoints/openai tests
.buildkite/test-pipeline.yaml
+5
@@ -319,6 +319,7 @@ steps:
319
- pytest -v -s compile/test_full_graph.py
320
321
- label: Kernels Core Operation Test
322
+ mirror_hardwares: [amd]
323
source_file_dependencies:
324
- csrc/
325
- tests/kernels/core
@@ -326,6 +327,7 @@ steps:
327
- pytest -v -s kernels/core
328
329
- label: Kernels Attention Test %N
330
+ mirror_hardwares: [amd]
331
source_file_dependencies:
332
- csrc/attention/
333
- vllm/attention
@@ -336,6 +338,7 @@ steps:
338
parallelism: 2
339
340
- label: Kernels Quantization Test %N
341
+ mirror_hardwares: [amd]
342
source_file_dependencies:
343
- csrc/quantization/
344
- vllm/model_executor/layers/quantization
@@ -345,6 +348,7 @@ steps:
348
parallelism: 2
349
350
- label: Kernels MoE Test
351
+ #mirror_hardwares: [amd]
352
source_file_dependencies:
353
- csrc/moe/
354
- tests/kernels/moe
@@ -353,6 +357,7 @@ steps:
357
- pytest -v -s kernels/moe
358
359
- label: Kernels Mamba Test
360
+ #mirror_hardwares: [amd]
361
source_file_dependencies:
362
- csrc/mamba/
363
- tests/kernels/mamba