Skip to content

Commit

Permalink
add missing file
Browse files Browse the repository at this point in the history
  • Loading branch information
laszewsk committed Oct 4, 2023
1 parent 145c0df commit b1d2772
Show file tree
Hide file tree
Showing 5 changed files with 410 additions and 1 deletion.
121 changes: 121 additions & 0 deletions benchmarks/cloudmask/target/greene_v0.5/a.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,121 @@
# config.yaml

# SciML-Bench
# Copyright © 2022 Scientific Machine Learning Research Group
# Scientific Computing Department, Rutherford Appleton Laboratory
# Science and Technology Facilities Council, UK.
# with modifications from Gregor von Laszewski, Juri Papay
# All rights reserved.

# This is a configuration file for the CloudMask benchmark.

name: cloudmask-greene


sbatch:
mode: h
dir: localscratch

run:
venvpath: "/scratch/{os.USER}/ENV3"
datadir: "/scratch/{os.USER}/data"
branch: 'main'
mode: "parallel"
# mode: "original"
# mode: "cloudmesh"

# Submission Information
submission:
benchmark: cloudmask
submitter: Gregor von Laszewski
email: [email protected]
org: University of Virginia
division: closed
version: mlcommons-cloudmask-v1.1
github_commit_version: TBD
status: completed
platform: rivanna
accelerators_per_node: 1

experiment:
# card_name: v100
card_name: v100
gpu_count: 1
cpu_num: 1
mem: "64GB"
repeat: "1"
epoch: 2
#epoch: 2
seed: 1234
learning_rate: 0.001
batch_size: 32
train_split: 0.8
clip_offset: 15
no_cache: False
nodes: 1
gpu: 1
early_stoppage_patience: "25"
early_stoppage: "False"

identifier: "{experiment.card_name}-{experiment.early_stoppage}-{experiment.early_stoppage_patience}-{experiment.epoch}-{experiment.learningrate}-{experiment.repeat}"

#system:
# host: "rivanna"
# python: "3.10.8"
# num_cpus: 1
# partition: "bii-gpu"
# # allocation: ds6011-sp22-002
# # allocation: bii_dsc
# allocation: bii_dsc_community
# reservation: bi_fox_dgx
# constraint: ""

system:
host: "greene"
python: "3.10.8"
num_cpus: 1
platform: greene

mask: "float"

# Training data
train_dir: "/scratch/{os.USER}/data/one-day"

# Inference data
inference_dir: "/scratch/{os.USER}/data/ssts"

# Model file
model_file: "{os.TARGET}/outputs/slstr_cloud/cloudModel-{identifier}.h5"

# training
training_loss: binary_crossentropy
training_metrics: accuracy


# Output directory
output_dir: "{os.TARGET}/outputs/slstr_cloud"

# Log file for recording runtimes
log_file: ./cloudmask_final_1.log

# Log file for MLCommons logging
mlperf_logfile: ./mlperf_cloudmask_final_1.log

# Size of each patch to feed to the network
PATCH_SIZE: 256

# Original height of the image
IMAGE_H: 1200

# Original width of the image
IMAGE_W: 1500

# No. of channels
N_CHANNELS: 9

# Min allowable SST
MIN_SST: 273.15

# Amount to crop the edges of the images by
CROP_SIZE: 80

121 changes: 121 additions & 0 deletions benchmarks/cloudmask/target/greene_v0.5/config_simple.1.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,121 @@
# config.yaml

# SciML-Bench
# Copyright © 2022 Scientific Machine Learning Research Group
# Scientific Computing Department, Rutherford Appleton Laboratory
# Science and Technology Facilities Council, UK.
# with modifications from Gregor von Laszewski, Juri Papay
# All rights reserved.

# This is a configuration file for the CloudMask benchmark.

name: cloudmask-greene

hyperparameter:
early_stoppage: False

sbatch:
mode: h
dir: localscratch

run:
venvpath: "/scratch/{os.USER}/ENV3"
datadir: ""/scratch/{os.USER}/data"
branch: 'main'
mode: "parallel"
# mode: "original"
# mode: "cloudmesh"

# Submission Information
submission:
benchmark: cloudmask
submitter: Gregor von Laszewski
email: [email protected]
org: University of Virginia
division: closed
version: mlcommons-cloudmask-v1.1
github_commit_version: TBD
status: completed
platform: rivanna
accelerators_per_node: 1

experiment:
# card_name: a100
card_name: v100
gpu_count: 1
cpu_num: 1
mem: "64GB"
repeat: "1"
epoch: 200
seed: 1234
learning_rate: 0.001
batch_size: 32
train_split: 0.8
clip_offset: 15
no_cache: False
nodes: 1
gpu: 1
early_stoppage_patience: 25



#system:
# host: "rivanna"
# python: "3.10.8"
# num_cpus: 1
# partition: "bii-gpu"
# # allocation: ds6011-sp22-002
# # allocation: bii_dsc
# allocation: bii_dsc_community
# reservation: bi_fox_dgx
# constraint: ""

system:
host: "greene"
python: "3.10.8"
num_cpus: 1
platform: greene

mask: "float"

# Training data
train_dir: "/scratch/{os.USER}/data/one-day"

# Inference data
inference_dir: "/scratch/{os.USER}/data/ssts"

# Model file
model_file: "/scratch/{os.USER}/github/mlcommons/benchmarks/cloudmask/target/greene_v0.5/outputs/slstr_cloud/cloudModel.h5"

# training
training_loss: binary_crossentropy
training_metrics: accuracy


# Output directory
output_dir: "/scratch/{os.USER}/github/mlcommons/benchmarks/cloudmask/target/greene_v0.5/outputs/slstr_cloud"

# Log file for recording runtimes
log_file: ./cloudmask_final_1.log

# Log file for MLCommons logging
mlperf_logfile: ./mlperf_cloudmask_final_1.log

# Size of each patch to feed to the network
PATCH_SIZE: 256

# Original height of the image
IMAGE_H: 1200

# Original width of the image
IMAGE_W: 1500

# No. of channels
N_CHANNELS: 9

# Min allowable SST
MIN_SST: 273.15

# Amount to crop the edges of the images by
CROP_SIZE: 80

122 changes: 122 additions & 0 deletions benchmarks/cloudmask/target/greene_v0.5/config_simple_rivanna.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,122 @@
# config.yaml

# SciML-Bench
# Copyright © 2022 Scientific Machine Learning Research Group
# Scientific Computing Department, Rutherford Appleton Laboratory
# Science and Technology Facilities Council, UK.
# with modifications from Gregor von Laszewski, Juri Papay
# All rights reserved.

# This is a configuration file for the CloudMask benchmark.

name: cloudmask-greene


sbatch:
mode: h
dir: localscratch

run:
venvpath: "/scratch/{os.USER}/ENV3"
datadir: "/scratch/{os.USER}/data"
branch: 'main'
mode: "parallel"
# mode: "original"
# mode: "cloudmesh"

# Submission Information
submission:
benchmark: cloudmask
submitter: Gregor von Laszewski
email: [email protected]
org: University of Virginia
division: closed
version: mlcommons-cloudmask-v1.1
github_commit_version: TBD
status: completed
platform: rivanna
accelerators_per_node: 1

experiment:
# card_name: v100
card_name: v100
gpu_count: 1
cpu_num: 1
mem: "64GB"
repeat: "1"
epoch: 2
#epoch: 2
seed: 1234
learning_rate: 0.001
batch_size: 32
train_split: 0.8
clip_offset: 15
no_cache: False
nodes: 1
gpu: 1
early_stoppage_patience: "25"
early_stoppage: "False"

identifier: "{experiment.card_name}-{experiment.early_stoppage}-{experiment.early_stoppage_patience}-{experiment.epoch}-{experiment.learningrate}-{experiment.repeat}"

#system:
# host: "rivanna"
# python: "3.10.8"
# num_cpus: 1
# partition: "bii-gpu"
# # allocation: ds6011-sp22-002
# # allocation: bii_dsc
# allocation: bii_dsc_community
# reservation: bi_fox_dgx
# constraint: ""

system:
host: "greene"
python: "3.10.8"
num_cpus: 1
platform: greene

mask: "float"

# Training data
train_dir: "/scratch/{os.USER}/data/one-day"

# Inference data
inference_dir: "/scratch/{os.USER}/data/ssts"


# training
training_loss: binary_crossentropy
training_metrics: accuracy


# Output directory
output_dir: "{os.TARGET}/outputs/slstr_cloud"

# Model file
model_file: "{output_dir}/cloudModel-{identifier}.h5"

# Log file for recording runtimes
log_file: "{output_dir}/cloudmask_final_{identifier}.log"

# Log file for MLCommons logging
mlperf_logfile: "{output_dir}/mlperf_cloudmask_final_{identifier}.log"

# Size of each patch to feed to the network
PATCH_SIZE: 256

# Original height of the image
IMAGE_H: 1200

# Original width of the image
IMAGE_W: 1500

# No. of channels
N_CHANNELS: 9

# Min allowable SST
MIN_SST: 273.15

# Amount to crop the edges of the images by
CROP_SIZE: 80

Loading

0 comments on commit b1d2772

Please sign in to comment.