Skip to content

Commit

Permalink
Remove resource limit from pytorch job (#466)
Browse files Browse the repository at this point in the history
* Remove resource limit from pytorch job

Signed-off-by: Yuvraj <code@evalsocket.dev>
  • Loading branch information
yindia authored Oct 29, 2021
1 parent 69ef957 commit a653b22
Show file tree
Hide file tree
Showing 3 changed files with 8 additions and 6 deletions.
2 changes: 1 addition & 1 deletion cookbook/integrations/kubernetes/kfmpi/requirements.in
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
flytekitplugins-kfmpi>=0.23.2
flytekit>=0.23.2
flytekit>=0.24.0b0
flyteidl>=0.21.4
tensorflow==2.6.0
8 changes: 5 additions & 3 deletions cookbook/integrations/kubernetes/kfmpi/requirements.txt
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,8 @@ clang==5.0
# via tensorflow
click==7.1.2
# via flytekit
cloudpickle==2.0.0
# via flytekit
croniter==1.0.15
# via flytekit
dataclasses-json==0.5.6
Expand All @@ -38,12 +40,12 @@ docstring-parser==0.12
# via flytekit
flatbuffers==1.12
# via tensorflow
flyteidl==0.21.7
flyteidl==0.21.8
# via
# -r requirements.in
# flytekit
# flytekitplugins-kfmpi
flytekit==0.23.2
flytekit==0.24.0b0
# via
# -r requirements.in
# flytekitplugins-kfmpi
Expand Down Expand Up @@ -106,7 +108,7 @@ opt-einsum==3.3.0
# via tensorflow
pandas==1.3.4
# via flytekit
protobuf==3.19.0
protobuf==3.19.1
# via
# flyteidl
# flytekit
Expand Down
4 changes: 2 additions & 2 deletions cookbook/integrations/kubernetes/kfpytorch/pytorch_mnist.py
Original file line number Diff line number Diff line change
Expand Up @@ -180,12 +180,12 @@ class Hyperparameters(object):
@task(
task_config=PyTorch(
num_workers=2,
per_replica_requests=Resources(cpu=cpu_request, mem=mem_request, gpu=gpu_request),
per_replica_limits=Resources(mem=mem_limit, gpu=gpu_limit),
),
retries=2,
cache=True,
cache_version="1.0",
requests=Resources(cpu=cpu_request, mem=mem_request, gpu=gpu_request),
limits=Resources(mem=mem_limit, gpu=gpu_limit)
)
def mnist_pytorch_job(hp: Hyperparameters) -> TrainingOutputs:
log_dir = "logs"
Expand Down

0 comments on commit a653b22

Please sign in to comment.