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
46847176
Commit
46847176
authored
4 years ago
by
Hashim Sharif
Browse files
Options
Downloads
Patches
Plain Diff
Adding measured network times over 4G LTE network
parent
b90abbe4
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/time_jetson_profiles.py
+33
-6
33 additions, 6 deletions
llvm/projects/hpvm-tensor-rt/bin/time_jetson_profiles.py
with
33 additions
and
6 deletions
llvm/projects/hpvm-tensor-rt/bin/time_jetson_profiles.py
+
33
−
6
View file @
46847176
...
...
@@ -19,7 +19,9 @@ ResNet50.binary_path = "resnet_imagenet"
ResNet50
.
binary_time
=
3.85
*
100
# 50 images * 100 batches
ResNet50
.
batch_time
=
3.85
# Time for batch with 50 images
ResNet50
.
num_layers
=
53
ResNet50
.
data_size
=
50
*
3
*
224
*
224
ResNet50
.
data_size
=
50
*
3
*
224
*
224
*
4
# *4 for Float32 Data
ResNet50
.
num_classes
=
1000
ResNet50
.
batch_size
=
50
...
...
@@ -29,8 +31,9 @@ VGG16_ImageNet.binary_path = "vgg16_imagenet"
VGG16_ImageNet
.
binary_time
=
4.55
*
100
# 50 images * 100 batches
VGG16_ImageNet
.
batch_time
=
4.55
VGG16_ImageNet
.
num_layers
=
16
VGG16_ImageNet
.
data_size
=
50
*
3
*
224
*
224
VGG16_ImageNet
.
data_size
=
50
*
3
*
224
*
224
*
4
VGG16_ImageNet
.
num_classes
=
1000
VGG16_ImageNet
.
batch_size
=
50
AlexNet_ImageNet
=
Benchmark
()
...
...
@@ -38,7 +41,9 @@ AlexNet_ImageNet.binary_path = "alexnet_imagenet"
AlexNet_ImageNet
.
binary_time
=
0.66
*
100
AlexNet_ImageNet
.
batch_time
=
0.66
AlexNet_ImageNet
.
num_layers
=
8
AlexNet_ImageNet
.
data_size
=
50
*
3
*
224
*
224
AlexNet_ImageNet
.
data_size
=
50
*
3
*
224
*
224
*
4
AlexNet_ImageNet
.
num_classes
=
1000
AlexNet_ImageNet
.
batch_size
=
50
...
...
@@ -51,6 +56,9 @@ promise_knobs = 7
total_machines
=
100
total_confs
=
50
download_time_per_1MB
=
(
6.1
*
60
)
/
100
# 6.1 mins over 4G LTE network for 100 MB data upload
upload_time_per_1MB
=
(
26.4
*
60
)
/
100
# 26.4 mins over 4G LTE network for 100 MB data upload
...
...
@@ -77,7 +85,7 @@ def getErrorProfileTime(Bench):
def
getConfTime
(
Bench
):
conf_per_machine
=
promise_conf_runs
*
(
total_confs
*
1.0
/
total_machines
)
conf_per_machine
=
promise_conf_runs
*
(
total_confs
*
1.0
/
total_machines
)
conf_time
=
conf_per_machine
*
Bench
.
binary_time
return
conf_time
...
...
@@ -85,13 +93,32 @@ def getConfTime(Bench):
def
getNetworkTime
(
Bench
):
# Calibration Download Time
download_data_MB
=
Bench
.
data_size
/
1000000
download_data_time
=
download_data_MB
*
download_time_per_1MB
# Profile Uploading (to Cloud Server) Time
total_knobs
=
(
promise_knobs
+
1
)
*
Bench
.
num_layers
profile_size
=
total_knobs
*
Bench
.
batch_size
*
Bench
.
num_classes
*
4
# *4 for FP32 data
profile_size_MB
=
profile_size
/
1000000
upload_data_time
=
profile_size_MB
*
upload_time_per_1MB
network_time
=
download_data_time
+
upload_data_time
return
network_time
def
getTimeOnEdge
(
Bench
):
err_time
=
getErrorProfileTime
(
Bench
)
conf_time
=
getConfTime
(
Bench
)
network_time
=
getNetworkTime
(
Bench
)
total_time
=
err_time
+
conf_time
total_time
=
err_time
+
conf_time
+
network_time
total_time
=
total_time
/
60
return
total_time
...
...
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