Releasing/1.8.4 extra (#15952)

* [App] Remove `SingleProcessRuntime` (#15933) * Remove SingleProcessRuntime * Remove unused queues * Docs (cherry picked from commit e250dfe) * [App] Fix bug when using structures with works (#15911) * Fix bug when using structures with works * Add test * Update CHANGELOG.md (cherry picked from commit 1283226) * [App] Wait for full file to be transferred in Path / Payload (#15934) * Wait for full file to be transferred in Path / Payload * Fixes (cherry picked from commit b8c7018) Co-authored-by: Ethan Harris <ethanwharris@gmail.com>
Lightning-AI · Dec 8, 2022 · e0e75d1 · e0e75d1
1 parent a0d3475
commit e0e75d1
Show file tree

Hide file tree

Showing 34 changed files with 160 additions and 231 deletions.
diff --git a/docs/source-app/api_reference/runners.rst b/docs/source-app/api_reference/runners.rst
@@ -18,5 +18,4 @@ ______________
     :template: classtemplate.rst
 
     ~cloud.CloudRuntime
-    ~singleprocess.SingleProcessRuntime
     ~multiprocess.MultiProcessRuntime
diff --git a/docs/source-app/api_references.rst b/docs/source-app/api_references.rst
@@ -89,7 +89,6 @@ _______
     :template: classtemplate_no_index.rst
 
     ~cloud.CloudRuntime
-    ~singleprocess.SingleProcessRuntime
     ~multiprocess.MultiProcessRuntime
 
 ----

diff --git a/docs/source-app/testing.rst b/docs/source-app/testing.rst
@@ -120,7 +120,6 @@ We provide ``application_testing`` as a helper funtion to get your application u
        os.path.join(_PROJECT_ROOT, "examples/app_v0/app.py"),
        "--blocking",
        "False",
-       "--multiprocess",
        "--open-ui",
        "False",
    ]
@@ -129,9 +128,7 @@ First in the list for ``command_line`` is the location of your script. It is an
 
 Next there are a couple of options you can leverage:
 
-
 * ``blocking`` - Blocking is an app status that says "Do not run until I click run in the UI". For our integration test, since we are not using the UI, we are setting this to "False".
-* ``multiprocess/singleprocess`` - This is the runtime your app is expected to run under.
 * ``open-ui`` - We set this to false since this is the routine that opens a browser for your local execution.
 
 Once you have your commandline ready, you will then be able to kick off the test and gather results:

diff --git a/examples/app_template_streamlit_ui/app.py b/examples/app_template_streamlit_ui/app.py
@@ -1,8 +1,8 @@
 import logging
 
-from lightning_app import LightningApp, LightningFlow
-from lightning_app.frontend import StreamlitFrontend
-from lightning_app.utilities.state import AppState
+from lightning.app import LightningApp, LightningFlow
+from lightning.app.frontend import StreamlitFrontend
+from lightning.app.utilities.state import AppState
 
 logger = logging.getLogger(__name__)
 
@@ -45,4 +45,4 @@ def configure_layout(self):
         return [{"name": "StreamLitUI", "content": self.streamlit_ui}]
 
 
-app = LightningApp(HelloWorld(), log_level="debug")
+app = LightningApp(HelloWorld())
diff --git a/pyproject.toml b/pyproject.toml
@@ -102,7 +102,6 @@ module = [
     "lightning_app.runners.cloud",
     "lightning_app.runners.multiprocess",
     "lightning_app.runners.runtime",
-    "lightning_app.runners.singleprocess",
     "lightning_app.source_code.copytree",
     "lightning_app.source_code.hashing",
     "lightning_app.source_code.local",

diff --git a/requirements/app/base.txt b/requirements/app/base.txt
@@ -7,7 +7,7 @@ fsspec>=2022.5.0, <=2022.7.1
 croniter>=1.3.0, <1.4.0  # strict; TODO: for now until we find something more robust.
 traitlets>=5.3.0, <=5.4.0
 arrow>=1.2.0, <1.2.4
-lightning-utilities>=0.3.*, !=0.4.0, <0.5.0
+lightning-utilities>=0.3.0, !=0.4.0, <0.5.0
 beautifulsoup4>=4.8.0, <4.11.2
 inquirer>=2.10.0
 psutil<5.9.4

diff --git a/requirements/lite/base.txt b/requirements/lite/base.txt
@@ -2,7 +2,7 @@
 #  in case you want to preserve/enforce restrictions on the latest compatible version, add "strict" as an in-line comment
 
 numpy>=1.17.2, <1.23.1
-torch>=1.9.*, <=1.13.0
+torch>=1.9.0, <=1.13.0
 fsspec[http]>2021.06.0, <2022.6.0
 packaging>=17.0, <=21.3
 typing-extensions>=4.0.0, <=4.4.0

diff --git a/requirements/lite/examples.txt b/requirements/lite/examples.txt
@@ -0,0 +1,4 @@
+# NOTE: the upper bound for the package version is only set for CI stability, and it is dropped while installing this package
+#  in case you want to preserve/enforce restrictions on the latest compatible version, add "strict" as an in-line comment
+
+torchvision>=0.10.0, <=0.13.0
diff --git a/requirements/pytorch/base.txt b/requirements/pytorch/base.txt
@@ -2,7 +2,7 @@
 #  in case you want to preserve/enforce restrictions on the latest compatible version, add "strict" as an in-line comment
 
 numpy>=1.17.2, <1.23.1
-torch>=1.9.*, <=1.13.0
+torch>=1.9.0, <=1.13.0
 tqdm>=4.57.0, <4.65.0
 PyYAML>=5.4, <=6.0
 fsspec[http]>2021.06.0, <2022.8.0

diff --git a/requirements/pytorch/examples.txt b/requirements/pytorch/examples.txt
@@ -1,6 +1,6 @@
 # NOTE: the upper bound for the package version is only set for CI stability, and it is dropped while installing this package
 #  in case you want to preserve/enforce restrictions on the latest compatible version, add "strict" as an in-line comment
 
-torchvision>=0.10.*, <=0.13.0
+torchvision>=0.10.0, <=0.14.0
 gym[classic_control]>=0.17.0, <0.26.3
 ipython[all] <8.6.1
diff --git a/src/lightning_app/CHANGELOG.md b/src/lightning_app/CHANGELOG.md
@@ -12,7 +12,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added the CLI command `lightning run model` to launch a `LightningLite` accelerated script ([#15506](https://github.com/Lightning-AI/lightning/pull/15506))
 - Added the CLI command `lightning delete app` to delete a lightning app on the cloud ([#15783](https://github.com/Lightning-AI/lightning/pull/15783))
 - Added a CloudMultiProcessBackend which enables running a child App from within the Flow in the cloud ([#15800](https://github.com/Lightning-AI/lightning/pull/15800))
-- Apps without UIs no longer activate the "Open App" button when running in the cloud ([#15875](https://github.com/Lightning-AI/lightning/pull/15875))
 - Added `AutoScaler` component ([#15769](https://github.com/Lightning-AI/lightning/pull/15769))
 - Added the property `ready` of the LightningFlow to inform when the `Open App` should be visible ([#15921](https://github.com/Lightning-AI/lightning/pull/15921))
 - Added private work attributed `_start_method` to customize how to start the works ([#15923](https://github.com/Lightning-AI/lightning/pull/15923))
@@ -26,6 +25,12 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Show a message when `BuildConfig(requirements=[...])` is passed but a `requirements.txt` file is already present in the Work ([#15799](https://github.com/Lightning-AI/lightning/pull/15799))
 - Show a message when `BuildConfig(dockerfile="...")` is passed but a `Dockerfile` file is already present in the Work ([#15799](https://github.com/Lightning-AI/lightning/pull/15799))
 - Dropped name column from cluster list ([#15721](https://github.com/Lightning-AI/lightning/pull/15721))
+- Apps without UIs no longer activate the "Open App" button when running in the cloud ([#15875](https://github.com/Lightning-AI/lightning/pull/15875))
+- Wait for full file to be transferred in Path / Payload ([#15934](https://github.com/Lightning-AI/lightning/pull/15934))
+
+### Removed
+
+- Removed the `SingleProcessRuntime` ([#15933](https://github.com/Lightning-AI/lightning/pull/15933))
 
 ### Fixed
 
@@ -37,6 +42,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed the `enable_spawn` method of the `WorkRunExecutor` ([#15812](https://github.com/Lightning-AI/lightning/pull/15812))
 - Fixed require/import decorator ([#15849](https://github.com/Lightning-AI/lightning/pull/15849))
 
+- Fixed a bug where using `L.app.structures` would cause multiple apps to be opened and fail with an error in the cloud ([#15911](https://github.com/Lightning-AI/lightning/pull/15911))
+
 
 ## [1.8.3] - 2022-11-22
 

diff --git a/src/lightning_app/components/auto_scaler.py b/src/lightning_app/components/auto_scaler.py
@@ -25,7 +25,6 @@
 from lightning_app.utilities.packaging.cloud_compute import CloudCompute
 
 logger = Logger(__name__)
-lock = asyncio.Lock()
 
 
 def _raise_granular_exception(exception: Exception) -> None:
@@ -209,6 +208,7 @@ async def process_request(self, data: BaseModel):
     def run(self):
 
         logger.info(f"servers: {self.servers}")
+        lock = asyncio.Lock()
 
         self._iter = cycle(self.servers)
         self._last_batch_sent = time.time()

diff --git a/src/lightning_app/core/app.py b/src/lightning_app/core/app.py
@@ -21,7 +21,7 @@
     FRONTEND_DIR,
     STATE_ACCUMULATE_WAIT,
 )
-from lightning_app.core.queues import BaseQueue, SingleProcessQueue
+from lightning_app.core.queues import BaseQueue
 from lightning_app.core.work import LightningWork
 from lightning_app.frontend import Frontend
 from lightning_app.storage import Drive, Path, Payload
@@ -549,8 +549,6 @@ def _collect_work_finish_status(self) -> dict:
     def _should_snapshot(self) -> bool:
         if len(self.works) == 0:
             return True
-        elif isinstance(self.delta_queue, SingleProcessQueue):
-            return True
         elif self._has_updated:
             work_finished_status = self._collect_work_finish_status()
             if work_finished_status:

diff --git a/src/lightning_app/core/constants.py b/src/lightning_app/core/constants.py
@@ -75,3 +75,7 @@ def get_lightning_cloud_url() -> str:
 
 def enable_multiple_works_in_default_container() -> bool:
     return bool(int(os.getenv("ENABLE_MULTIPLE_WORKS_IN_DEFAULT_CONTAINER", "0")))
+
+
+# Number of seconds to wait between filesystem checks when waiting for files in remote storage
+REMOTE_STORAGE_WAIT = 0.5
diff --git a/src/lightning_app/core/queues.py b/src/lightning_app/core/queues.py
@@ -49,7 +49,6 @@
 
 
 class QueuingSystem(Enum):
-    SINGLEPROCESS = "singleprocess"
     MULTIPROCESS = "multiprocess"
     REDIS = "redis"
     HTTP = "http"
@@ -59,10 +58,8 @@ def get_queue(self, queue_name: str) -> "BaseQueue":
             return MultiProcessQueue(queue_name, default_timeout=STATE_UPDATE_TIMEOUT)
         elif self == QueuingSystem.REDIS:
             return RedisQueue(queue_name, default_timeout=REDIS_QUEUES_READ_DEFAULT_TIMEOUT)
-        elif self == QueuingSystem.HTTP:
-            return HTTPQueue(queue_name, default_timeout=STATE_UPDATE_TIMEOUT)
         else:
-            return SingleProcessQueue(queue_name, default_timeout=STATE_UPDATE_TIMEOUT)
+            return HTTPQueue(queue_name, default_timeout=STATE_UPDATE_TIMEOUT)
 
     def get_api_response_queue(self, queue_id: Optional[str] = None) -> "BaseQueue":
         queue_name = f"{queue_id}_{API_RESPONSE_QUEUE_CONSTANT}" if queue_id else API_RESPONSE_QUEUE_CONSTANT
@@ -179,21 +176,6 @@ def is_running(self) -> bool:
         return True
 
 
-class SingleProcessQueue(BaseQueue):
-    def __init__(self, name: str, default_timeout: float):
-        self.name = name
-        self.default_timeout = default_timeout
-        self.queue = queue.Queue()
-
-    def put(self, item):
-        self.queue.put(item)
-
-    def get(self, timeout: int = None):
-        if timeout == 0:
-            timeout = self.default_timeout
-        return self.queue.get(timeout=timeout, block=(timeout is None))
-
-
 class MultiProcessQueue(BaseQueue):
     def __init__(self, name: str, default_timeout: float):
         self.name = name

diff --git a/src/lightning_app/runners/__init__.py b/src/lightning_app/runners/__init__.py
@@ -1,7 +1,6 @@
 from lightning_app.runners.cloud import CloudRuntime
 from lightning_app.runners.multiprocess import MultiProcessRuntime
 from lightning_app.runners.runtime import dispatch, Runtime
-from lightning_app.runners.singleprocess import SingleProcessRuntime
 from lightning_app.utilities.app_commands import run_app_commands
 from lightning_app.utilities.load_app import load_app_from_file
 
@@ -11,6 +10,5 @@
     "run_app_commands",
     "Runtime",
     "MultiProcessRuntime",
-    "SingleProcessRuntime",
     "CloudRuntime",
 ]
diff --git a/src/lightning_app/runners/runtime_type.py b/src/lightning_app/runners/runtime_type.py
@@ -1,21 +1,18 @@
 from enum import Enum
 from typing import Type, TYPE_CHECKING
 
-from lightning_app.runners import CloudRuntime, MultiProcessRuntime, SingleProcessRuntime
+from lightning_app.runners import CloudRuntime, MultiProcessRuntime
 
 if TYPE_CHECKING:
     from lightning_app.runners.runtime import Runtime
 
 
 class RuntimeType(Enum):
-    SINGLEPROCESS = "singleprocess"
     MULTIPROCESS = "multiprocess"
     CLOUD = "cloud"
 
     def get_runtime(self) -> Type["Runtime"]:
-        if self == RuntimeType.SINGLEPROCESS:
-            return SingleProcessRuntime
-        elif self == RuntimeType.MULTIPROCESS:
+        if self == RuntimeType.MULTIPROCESS:
             return MultiProcessRuntime
         elif self == RuntimeType.CLOUD:
             return CloudRuntime

diff --git a/src/lightning_app/runners/singleprocess.py b/src/lightning_app/runners/singleprocess.py
diff --git a/src/lightning_app/storage/orchestrator.py b/src/lightning_app/storage/orchestrator.py
@@ -105,6 +105,7 @@ def run_once(self, work_name: str) -> None:
                             name=request.name,
                             path=maybe_artifact_path,
                             hash=request.hash,
+                            size=self.fs.info(maybe_artifact_path)["size"],
                             destination=request.destination,
                         )
                     if isinstance(request, _ExistsRequest):
@@ -139,6 +140,7 @@ def run_once(self, work_name: str) -> None:
                             path=request.path,
                             name=request.name,
                             hash=request.hash,
+                            size=0,
                             destination=request.destination,
                         )
                     if isinstance(request, _ExistsRequest):

diff --git a/src/lightning_app/storage/path.py b/src/lightning_app/storage/path.py
@@ -10,6 +10,7 @@
 from fsspec import AbstractFileSystem
 from fsspec.implementations.local import LocalFileSystem
 
+from lightning_app.core.constants import REMOTE_STORAGE_WAIT
 from lightning_app.core.queues import BaseQueue
 from lightning_app.storage.requests import _ExistsRequest, _ExistsResponse, _GetRequest, _GetResponse
 from lightning_app.utilities.app_helpers import Logger
@@ -199,9 +200,8 @@ def get(self, overwrite: bool = False) -> None:
         fs = _filesystem()
 
         # 3. Wait until the file appears in shared storage
-        while not fs.exists(response.path):
-            # TODO: Existence check on folder is not enough, files may not be completely transferred yet
-            sleep(0.5)
+        while not fs.exists(response.path) or fs.info(response.path)["size"] != response.size:
+            sleep(REMOTE_STORAGE_WAIT)
 
         if self.exists_local() and self.is_dir():
             # Delete the directory, otherwise we can't overwrite it
@@ -340,10 +340,11 @@ def _handle_get_request(work: "LightningWork", request: _GetRequest) -> _GetResp
         destination_path = _shared_storage_path() / request.hash
         response = _GetResponse(
             source=request.source,
+            name=request.name,
             path=str(destination_path),
             hash=request.hash,
+            size=source_path.stat().st_size,
             destination=request.destination,
-            name=request.name,
         )
 
         try:

diff --git a/src/lightning_app/storage/payload.py b/src/lightning_app/storage/payload.py
@@ -5,6 +5,7 @@
 from time import sleep
 from typing import Any, Optional, TYPE_CHECKING, Union
 
+from lightning_app.core.constants import REMOTE_STORAGE_WAIT
 from lightning_app.core.queues import BaseQueue
 from lightning_app.storage.path import _filesystem, _shared_storage_path, Path
 from lightning_app.storage.requests import _ExistsRequest, _ExistsResponse, _GetRequest, _GetResponse
@@ -159,9 +160,8 @@ def get(self) -> Any:
         fs = _filesystem()
 
         # 3. Wait until the file appears in shared storage
-        while not fs.exists(response.path):
-            # TODO: Existence check on folder is not enough, files may not be completely transferred yet
-            sleep(0.5)
+        while not fs.exists(response.path) or fs.info(response.path)["size"] != response.size:
+            sleep(REMOTE_STORAGE_WAIT)
 
         # 4. Copy the file from the shared storage to the destination on the local filesystem
         local_path = self._path
@@ -234,6 +234,7 @@ def _handle_get_request(work: "LightningWork", request: _GetRequest) -> _GetResp
         try:
             payload = getattr(work, request.name)
             payload.save(payload.value, source_path)
+            response.size = source_path.stat().st_size
             _copy_files(source_path, destination_path)
             _logger.debug(f"All files copied from {request.path} to {response.path}.")
         except Exception as e:

diff --git a/src/lightning_app/storage/requests.py b/src/lightning_app/storage/requests.py
@@ -17,6 +17,7 @@ class _GetResponse:
     name: str
     path: str
     hash: str
+    size: int = 0
     destination: str = ""
     exception: Optional[Exception] = None
     timedelta: Optional[float] = None