Revert "[CI] Add comprehensive testing: migration, e2e, and bench (#30)"

This reverts commit 91a6454.

.github/workflows/pylint.yml

@@ -1,4 +1,4 @@
-name: pylint
+name: Pylint
 
 on:
   push:
@@ -9,24 +9,21 @@ on:
     - main
 
 jobs:
-  cancel_previous_workflows:
-    runs-on: [self-hosted]
-    timeout-minutes: 3
-    steps:
-    - uses: styfle/[email protected]
-      with:
-        all_but_latest: true
-
-  pylint_test:
-    needs: cancel_previous_workflows
-    runs-on: [self-hosted]
-    timeout-minutes: 10
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.8", "3.9", "3.10"]
     steps:
     - uses: actions/checkout@v4
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v3
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install pylint==2.12.2
     - name: Analysing the code with pylint
       run: |
-        nvidia-docker run --rm -t --net host --ipc host \
-          -v ${PWD}:/workspace \
-          -w /workspace \
-          registry.cn-beijing.aliyuncs.com/llumnix/llumnix-dev:20240909_action_678a439 \
-          bash -c "pip install -e . > /dev/null && make lint"
+        pylint --rcfile=.pylintrc --output-format=parseable --jobs=8 $( find llumnix/ -type f -name '*.py')

Makefile

@@ -21,15 +21,15 @@ install:
 
 .PHONY: lint
 lint: check_pylint_installed check_pytest_installed
-	@pylint --rcfile=.pylintrc -s n  --jobs=32 ./llumnix
+	@pylint --rcfile=.pylintrc -s n ./llumnix
 
 	@pylint --rcfile=.pylintrc \
 			--disable=protected-access,super-init-not-called,unused-argument,redefined-outer-name,invalid-name \
-			-s n --jobs=32 ./tests
+			-s n ./tests
 
 .PHONY: test
 test: check_pytest_installed
-	@pytest -x -v --ignore=third_party/ --ignore=tests/e2e_test --disable-warnings
+	@pytest -x -q --ignore=third_party/ --disable-warnings
 
 #################### pygloo install for gloo migration backend begin ####################
 

llumnix/backends/vllm/llm_engine.py

@@ -169,7 +169,6 @@ def step(self) -> None:
         instance_info.step_id = next(self.step_counter)
         instance_info.timestamp = time.time()
         instance_info.latency = self.model_executor.last_inference_latency
-
         seq_groups = self.scheduler.running
         if seq_groups:
             tot_blocks = []
@@ -261,8 +260,8 @@ def commit_dst_request(self, backend_request: SequenceGroupLlumnix) -> None:
         logger.info("add seq {} to block table".format(seq.seq_id))
         pre_alloc_blocks = self.engine.scheduler.pre_alloc_cache_dict.pop(backend_request.request_id)
         self.engine.scheduler.block_manager.add_block_table(pre_alloc_blocks, seq.seq_id)
-        backend_request.reset_migration_args()
         self.add_running_request(backend_request)
+        backend_request.reset_migration_args()
 
     def send_blocks(self, dst_ray_actor: "ray.actor.ActorHandle", src_blocks: List[int], dst_blocks: List[int]) -> None:
         ray.get(dst_ray_actor.execute_engine_method.remote("_run_workers",

llumnix/backends/vllm/worker.py

@@ -11,7 +11,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import time
 from typing import Dict, List
 import math
 import ray
@@ -22,13 +21,11 @@
 from vllm.worker.worker import Worker
 from vllm.config import CacheConfig,  ModelConfig, ParallelConfig
 from vllm.worker.cache_engine import CacheEngine
-from vllm.config import _GB
 
 from llumnix.logger import init_logger
 from llumnix.backends.vllm.utils import _sample_with_torch
 from llumnix.backends.vllm.migration_backend import MigrationBackendBase, get_migration_backend
 from llumnix.internal_config import MigrationConfig
-from llumnix.utils import convert_bytes
 
 logger = init_logger(__name__)
 
@@ -107,19 +104,10 @@ def init_migration(self, instance_id: str, migration_config: MigrationConfig, sr
 
     def migrate_cache(self, src_worker_handle_list, src_blocks: List[int], dst_blocks: List[int]) -> None:
         src_worker_handle = src_worker_handle_list[self.rank]
-
-        start_time = time.time()
         try:
             self.migration_backend.migrate_cache(src_worker_handle, src_blocks, dst_blocks)
         except ray.exceptions.RayActorError:
             logger.info("[migrate_cache] self.rank: {}, src_worker_handle {} is dead".format(self.rank, src_worker_handle))
-        end_time = time.time()
-
-        total_kv_cache_size = len(src_blocks) * CacheEngine.get_cache_block_size(
-            self.cache_config, self.model_config, self.parallel_config)
-        speed = total_kv_cache_size/_GB/(end_time - start_time)
-        logger.info("[migration_cache] blocks_num: {}, total_kv_cache_size: {}, time: {}s, speed: {}GB/s."
-                    .format(len(src_blocks), convert_bytes(total_kv_cache_size), end_time-start_time, speed))
 
     def do_recv(self, *args, **kwargs):
         return self.migration_backend.do_recv(*args, **kwargs)

llumnix/entrypoints/llumnix_utils.py

@@ -17,7 +17,6 @@
 import time
 from typing import List, Tuple
 import asyncio
-import socket
 import ray
 
 from llumnix.llm_engine_manager import LLMEngineManager, MANAGER_ACTOR_NAME
@@ -39,9 +38,10 @@
 MAX_TASK_RETRIES = 300
 RETRIES_INTERVALS = 0.1
 
+
 def get_ip_address():
-    hostname = socket.gethostname()
-    ip_address = socket.gethostbyname(hostname)
+    result = subprocess.run(['hostname', '-i'], stdout=subprocess.PIPE, check=True)
+    ip_address = result.stdout.decode('utf-8').strip()
     return ip_address
 
 def launch_ray_cluster(ray_cluster_port: int) -> subprocess.CompletedProcess:

llumnix/utils.py

@@ -16,17 +16,3 @@
 
 def random_uuid() -> str:
     return str(uuid.uuid4().hex)
-
-def convert_bytes(bytes_size):
-    """Convert bytes to KB, MB, GB, etc."""
-    if bytes_size < 0:
-        raise ValueError("Size must be a non-negative integer.")
-
-    size_suffixes = ['B', 'KB', 'MB', 'GB', 'TB']
-    index = 0
-
-    while bytes_size >= 1024 and index < len(size_suffixes) - 1:
-        bytes_size /= 1024.0
-        index += 1
-
-    return f"{bytes_size:.2f} {size_suffixes[index]}"

pytest.ini

@@ -1,2 +1,2 @@
 [pytest]
-asyncio_default_fixture_loop_scope = function
+asyncio_default_fixture_loop_scope = function

tests/unit_test/backends/vllm/test_migration.py → tests/backends/vllm/test_migration.py

@@ -24,10 +24,9 @@
 from llumnix.internal_config import MigrationConfig
 from llumnix.llumlet.request import LlumnixRequest, RequestInferenceType
 
-# pylint: disable=unused-import
-from tests.unit_test.rpc.test_queue import request_output_queue_server
-# pylint: disable=unused-import
-from tests.conftest import setup_ray_env
+from tests.utils import setup_ray_env
+from tests.rpc.test_queue import request_output_queue_server
+
 
 from .test_llm_engine import MockEngine
 from .utils import create_dummy_prompt