Rooba
/
lightning
mirror of https://github.com/Lightning-AI/lightning.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

[pre-commit.ci] auto fixes from pre-commit.com hooks 

for more information, see https://pre-commit.ci
This commit is contained in:
pre-commit-ci[bot] 2022-12-05 16:20:41 +00:00
parent 56ea78b45f
commit 24983a0a5a
1 changed files with 52 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

104
src/lightning_app/components/auto_scaler.py
View File

@ -340,63 +340,63 @@ class _LoadBalancer(LightningWork):
class AutoScaler(LightningFlow): class AutoScaler(LightningFlow):
"""The ``AutoScaler`` can be used to automatically change the number of replicas of the given server in """The ``AutoScaler`` can be used to automatically change the number of replicas of the given server in
response to changes in the number of incoming requests. Incoming requests will be batched and balanced across response to changes in the number of incoming requests. Incoming requests will be batched and balanced across
the replicas. the replicas.
Note that the ``Autoscaler`` experience on the cloud is in beta. Note that the ``Autoscaler`` experience on the cloud is in beta.
Args: Args:
min_replicas: The number of works to start when app initializes. min_replicas: The number of works to start when app initializes.
max_replicas: The max number of works to spawn to handle the incoming requests. max_replicas: The max number of works to spawn to handle the incoming requests.
autoscale_interval: The number of seconds to wait before checking whether to upscale or downscale the works. autoscale_interval: The number of seconds to wait before checking whether to upscale or downscale the works.
endpoint: Default=api/predict. Provide the REST API path endpoint: Default=api/predict. Provide the REST API path
max_batch_size: (auto-batching) The number of requests to process at once. max_batch_size: (auto-batching) The number of requests to process at once.
timeout_batching: (auto-batching) The number of seconds to wait before sending the requests to process. timeout_batching: (auto-batching) The number of seconds to wait before sending the requests to process.
input_type: Input type. input_type: Input type.
output_type: Output type. output_type: Output type.
.. testcode:: .. testcode::
import lightning as L import lightning as L
# Example 1: Auto-scaling serve component out-of-the-box # Example 1: Auto-scaling serve component out-of-the-box
app = L.LightningApp( app = L.LightningApp(
L.app.components.AutoScaler( L.app.components.AutoScaler(
MyPythonServer, MyPythonServer,
min_replicas=1, min_replicas=1,
max_replicas=8, max_replicas=8,
autoscale_interval=10, autoscale_interval=10,
) )
) )
# Example 2: Customizing the scaling logic # Example 2: Customizing the scaling logic
class MyAutoScaler(L.app.components.AutoScaler): class MyAutoScaler(L.app.components.AutoScaler):
def scale(self, replicas: int, metrics: dict) -> int: def scale(self, replicas: int, metrics: dict) -> int:
pending_requests_per_running_or_pending_work = metrics["pending_requests"] / ( pending_requests_per_running_or_pending_work = metrics["pending_requests"] / (
replicas + metrics["pending_works"] replicas + metrics["pending_works"]
)
# upscale
max_requests_per_work = self.max_batch_size
if pending_requests_per_running_or_pending_work >= max_requests_per_work:
return replicas + 1
# downscale
min_requests_per_work = max_requests_per_work * 0.25
if pending_requests_per_running_or_pending_work < min_requests_per_work:
return replicas - 1
return replicas
app = L.LightningApp(
MyAutoScaler(
MyPythonServer,
min_replicas=1,
max_replicas=8,
autoscale_interval=10,
max_batch_size=8, # for auto batching
timeout_batching=2, # for auto batching
) )
# upscale
max_requests_per_work = self.max_batch_size
if pending_requests_per_running_or_pending_work >= max_requests_per_work:
return replicas + 1
# downscale
min_requests_per_work = max_requests_per_work * 0.25
if pending_requests_per_running_or_pending_work < min_requests_per_work:
return replicas - 1
return replicas
app = L.LightningApp(
MyAutoScaler(
MyPythonServer,
min_replicas=1,
max_replicas=8,
autoscale_interval=10,
max_batch_size=8, # for auto batching
timeout_batching=2, # for auto batching
) )
)
""" """
def __init__( def __init__(