Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
H
hpvm-release
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Model registry
Operate
Environments
Monitor
Incidents
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
llvm
hpvm-release
Commits
5281cb3a
Commit
5281cb3a
authored
4 years ago
by
Hashim Sharif
Browse files
Options
Downloads
Patches
Plain Diff
Updating install time computation script with times with 150 OpenTuner runs
parent
5f2378a3
No related branches found
Branches containing commit
No related tags found
Tags containing commit
No related merge requests found
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
llvm/projects/hpvm-tensor-rt/bin/compute_install_times.py
+31
-30
31 additions, 30 deletions
llvm/projects/hpvm-tensor-rt/bin/compute_install_times.py
with
31 additions
and
30 deletions
llvm/projects/hpvm-tensor-rt/bin/compute_install_times.py
+
31
−
30
View file @
5281cb3a
class
TuningParameters
:
def
__init__
(
self
):
self
.
iterations_measured
=
150
self
.
total_iterations
=
30000
tunerParams
=
TuningParameters
()
class
Benchmark
:
def
__init__
(
self
):
...
...
@@ -7,45 +15,44 @@ class Benchmark:
### All times are real profiled times on the Jetson Board
### Times are for 50
00 Images
### Times are for
1
50
OpenTuner iterations on Jetson
ResNet50
=
Benchmark
()
ResNet50
.
binary_time
=
3.85
*
100
# 50 images * 100 batches
ResNet18
=
Benchmark
()
ResNet18
.
binary_time
=
12.9
# 12.9 measured for 1000 images
MobileNet
=
Benchmark
()
MobileNet
.
binary_time
=
21.6
# 50 images * 100 batches
ResNet50
.
tuner_time
=
3.85
*
100
*
150
# 50 images * 100 batches
VGG16_ImageNet
=
Benchmark
()
VGG16_ImageNet
.
binary_time
=
4.55
*
100
# 50 images * 100 batches
VGG16_ImageNet
.
tuner_time
=
4.55
*
100
*
150
# 50 images * 100 batches
AlexNet_ImageNet
=
Benchmark
()
AlexNet_ImageNet
.
tuner_time
=
0.7
*
100
*
150
VGG16_CIFAR10
=
Benchmark
()
VGG16_CIFAR10
.
binary
_time
=
1
8.9
# 50 images * 100 batches
VGG16_CIFAR10
.
tuner
_time
=
1
.54
*
60
*
60
# 50 images * 100 batches
VGG16_CIFAR100
=
Benchmark
()
VGG16_CIFAR100
.
binary
_time
=
1
8.9
# 50 images * 100 batches
VGG16_CIFAR100
.
tuner
_time
=
1
.57
*
60
*
60
# 50 images * 100 batches
AlexNet_ImageNet
=
Benchmark
()
AlexNet_ImageNet
.
binary_time
=
0.7
*
100
ResNet18
=
Benchmark
()
ResNet18
.
tuner_time
=
0.52
*
60
*
60
# 12.9 measured for 1000 images
MobileNet
=
Benchmark
()
MobileNet
.
tuner_time
=
0.72
*
60
*
60
# 50 images * 100 batches
AlexNet_CIFAR10
=
Benchmark
()
AlexNet_CIFAR10
.
binary_time
=
13.5
AlexNet_CIFAR10
.
tuner_time
=
0.67
*
60
*
60
# Time in hours
AlexNet2_CIFAR10
=
Benchmark
()
AlexNet2_CIFAR10
.
binary
_time
=
5.56
AlexNet2_CIFAR10
.
tuner
_time
=
0.19
*
60
*
60
LeNet_CIFAR10
=
Benchmark
()
LeNet_CIFAR10
.
binary
_time
=
1.22
LeNet_CIFAR10
.
tuner
_time
=
0.11
*
60
*
60
...
...
@@ -54,18 +61,12 @@ LeNet_CIFAR10.binary_time = 1.22
def
getInstallTime
(
Bench
):
## We limit pareto configs to 50 after iterations of tuning complete
pareto_configs
=
50
promise_runs
=
30
iterations
=
10000
thresholds
=
3
total_runs
=
iterations
*
thresholds
t
otal_iteration_time
=
total_runs
*
Bench
.
binary_time
total_validation_time
=
pareto_configs
*
promise_runs
*
Bench
.
binary_time
total_time
=
total_iteration_time
+
total_validation_time
time_hours
=
total
_time
/
(
60
*
60
)
t
uner_invocations
=
tunerParams
.
total_iterations
/
tunerParams
.
iterations_measured
extrapolated_time
=
tuner_invocations
*
Bench
.
tuner_time
time_hours
=
extrapolated
_time
/
(
60
*
60
)
return
time_hours
...
...
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment