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 17, 00:30 UTC
[CI] Disable flash_attn backend for spec decode (#5286)
Simon Mo committed
Jun 5, 2024 at 17:49 UTC
3a6ae1d33c7a8ef28b6dfa978f53b4fdcdbaaca6
1 file changed
+5
-2
.buildkite/test-pipeline.yaml
+5
-2
@@ -45,7 +45,7 @@ steps:
45
- TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_basic_distributed_correctness.py
46
- TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_chunked_prefill_distributed.py
47
- TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_chunked_prefill_distributed.py
48
- - pytest -v -s spec_decode/e2e/test_integration_dist.py
48
+ - pytest -v -s spec_decode/e2e/test_integration_dist.py
49
50
- label: Distributed Tests (Multiple Groups)
51
#mirror_hardwares: [amd]
@@ -124,7 +124,10 @@ steps:
124
125
- label: Speculative decoding tests
126
#mirror_hardwares: [amd]
127
- command: pytest -v -s spec_decode
127
+ commands:
128
+ # See https://github.com/vllm-project/vllm/issues/5152
129
+ - export VLLM_ATTENTION_BACKEND=XFORMERS
130
+ - pytest -v -s spec_decode
131
132
- label: LoRA Test %N
133
#mirror_hardwares: [amd]