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 16, 10:45 UTC
mypy: check additional directories (#9162)
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Russell Bryant committed
Oct 8, 2024 at 18:08 UTC
2a131965a8144d571a4a211a44d1fc32e202ae10
4 files changed
+41
-26
.github/workflows/mypy.yaml
+2
-13
@@ -11,7 +11,7 @@ on:
11
- main
12
13
jobs:
14
- ruff:
14
+ mypy:
15
runs-on: ubuntu-latest
16
strategy:
17
matrix:
@@ -32,15 +32,4 @@ jobs:
32
pip install types-setuptools
33
- name: Mypy
34
run: |
35
- mypy
36
- mypy tests --follow-imports skip
37
- mypy vllm/attention --follow-imports skip
38
- mypy vllm/distributed --follow-imports skip
39
- mypy vllm/engine --follow-imports skip
40
- mypy vllm/executor --follow-imports skip
41
- mypy vllm/lora --follow-imports skip
42
- mypy vllm/model_executor --follow-imports skip
43
- mypy vllm/prompt_adapter --follow-imports skip
44
- mypy vllm/spec_decode --follow-imports skip
45
- mypy vllm/worker --follow-imports skip
46
-
35
+ tools/mypy.sh
format.sh
+1
-11
@@ -96,17 +96,7 @@ echo 'vLLM yapf: Done'
96
97
# Run mypy
98
echo 'vLLM mypy:'
99
-mypy --follow-imports skip # Note that this is less strict than CI
100
-mypy tests --follow-imports skip
101
-mypy vllm/attention --follow-imports skip
102
-mypy vllm/distributed --follow-imports skip
103
-mypy vllm/engine --follow-imports skip
104
-mypy vllm/executor --follow-imports skip
105
-mypy vllm/lora --follow-imports skip
106
-mypy vllm/model_executor --follow-imports skip
107
-mypy vllm/prompt_adapter --follow-imports skip
108
-mypy vllm/spec_decode --follow-imports skip
109
-mypy vllm/worker --follow-imports skip
99
+tools/mypy.sh
100
echo 'vLLM mypy: Done'
101
102
tools/mypy.sh
new
+36
@@ -0,0 +1,36 @@
1
+#!/bin/bash
2
+
3
+CI=${1:-0}
4
+
5
+run_mypy() {
6
+ echo "Running mypy on $1"
7
+ if [ $CI -eq 1 ] && [ -z "$1" ]; then
8
+ mypy "$@"
9
+ return
10
+ fi
11
+ mypy --follow-imports skip "$@"
12
+}
13
+
14
+run_mypy # Note that this is less strict than CI
15
+run_mypy tests
16
+run_mypy vllm/assets
17
+run_mypy vllm/attention
18
+#run_mypy vllm/compilation
19
+#run_mypy vllm/core
20
+run_mypy vllm/distributed
21
+run_mypy vllm/engine
22
+#run_mypy vllm/entrypoints
23
+run_mypy vllm/executor
24
+#run_mypy vllm/inputs
25
+run_mypy vllm/logging
26
+run_mypy vllm/lora
27
+run_mypy vllm/model_executor
28
+run_mypy vllm/multimodal
29
+run_mypy vllm/platforms
30
+run_mypy vllm/plugins
31
+run_mypy vllm/prompt_adapter
32
+run_mypy vllm/spec_decode
33
+run_mypy vllm/transformers_utils
34
+run_mypy vllm/usage
35
+#run_mypy vllm/vllm_flash_attn
36
+run_mypy vllm/worker
vllm/usage/usage_lib.py
+2
-2
@@ -204,7 +204,7 @@ class UsageMessage:
204
self._write_to_file(data)
205
self._send_to_server(data)
206
207
- def _send_to_server(self, data):
207
+ def _send_to_server(self, data: Dict[str, Any]) -> None:
208
try:
209
global_http_client = global_http_connection.get_sync_client()
210
global_http_client.post(_USAGE_STATS_SERVER, json=data)
@@ -212,7 +212,7 @@ class UsageMessage:
212
# silently ignore unless we are using debug log
213
logging.debug("Failed to send usage data to server")
214
215
- def _write_to_file(self, data):
215
+ def _write_to_file(self, data: Dict[str, Any]) -> None:
216
os.makedirs(os.path.dirname(_USAGE_STATS_JSON_PATH), exist_ok=True)
217
Path(_USAGE_STATS_JSON_PATH).touch(exist_ok=True)
218
with open(_USAGE_STATS_JSON_PATH, "a") as f: