@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 93348d9458
Fork 0

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, 04:15 UTC

[CI] Shard tests for LoRA and Kernels to speed up (#3445)

Simon Mo committed Mar 17, 2024 at 14:56 UTC 93348d9458af7517bb8c114611d438a1b4a2c3be
3 files changed +10 -5
.buildkite/test-pipeline.yaml
+6 -5
@@ -33,9 +33,9 @@ steps:
33 - label: Entrypoints Test
34 command: pytest -v -s entrypoints
35
36 -- label: Kernels Test
37 - command: pytest -v -s kernels
38 - soft_fail: true
36 +- label: Kernels Test %N
37 + command: pytest -v -s kernels --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
38 + parallelism: 4
39
40 - label: Models Test
41 commands:
@@ -55,8 +55,9 @@ steps:
55 - label: Speculative decoding tests
56 command: pytest -v -s spec_decode
57
58 -- label: LoRA Test
59 - command: pytest -v -s lora --forked
58 +- label: LoRA Test %N
59 + command: pytest -v -s lora --forked --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
60 + parallelism: 4
61
62 - label: Metrics Test
63 command: pytest -v -s metrics
.buildkite/test-template.j2
+3
@@ -20,6 +20,9 @@ steps:
20 agents:
21 queue: kubernetes
22 soft_fail: {{ step.soft_fail or false }}
23 + {% if step.parallelism %}
24 + parallelism: {{ step.parallelism }}
25 + {% endif %}
26 retry:
27 automatic:
28 - exit_status: -1 # Agent was lost
requirements-dev.txt
+1
@@ -16,6 +16,7 @@ pytest
16 pytest-forked
17 pytest-asyncio
18 pytest-rerunfailures
19 +pytest-shard
20 httpx
21 einops # required for MPT
22 openai