Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[SD-94] Ablation study for resnet18 #44

Draft
wants to merge 67 commits into
base: develop
Choose a base branch
from
Draft

[SD-94] Ablation study for resnet18 #44

wants to merge 67 commits into from

Conversation

d-lowl
Copy link
Member

@d-lowl d-lowl commented Jan 10, 2025 •
edited
Loading

  • Add code for layer ablation: probe the output shape, and return random tensor of that shape
  • Add a script to detect which layers to ablate for the study
  • Add raw data collected with ablated networks
  • Add exploratory data analysis notebook for the said data

d-lowl and others added 30 commits January 2, 2025 14:55
@d-lowl
Add methods for ablation of resnet18
8755783
@d-lowl
Pad and reshape instead of replacing with ones
be23d31
@d-lowl
Add optional ablation to the inference power script
f582fc9
@d-lowl
Add log line for ablation
11ce263
@d-lowl
Log ablated models separately
efca577
@d-lowl
Add all layers for ablation in the run experiment script
f48bfce
@d-lowl
Use correct layer detection script
c5637c3
@d-lowl
Use correct layer detection script
165fd1f
@d-lowl
Use correct layer detection script
a3525a3
@d-lowl
Use correct layer detection script
37c3622
@osw282
Add resnet18 ablated results
e022fa2
@d-lowl
Add notebook for EDA of raw data after ablation
d548f5d
@d-lowl
Pass zeros instead of actual reshaping
aadf3b0
@d-lowl
Pass zeros instead of actual reshaping
0e845af
@d-lowl
Push zeros to cuda correctly?
61223eb
@d-lowl
revert back to ablation with padding
3f39b4c
@d-lowl
Temprorarily skip already measured layers
f53cb24
@d-lowl
Temprorarily skip already measured layers
e2d5cd7
@d-lowl
Temprorarily skip already measured layers
a16b13e
@osw282
Collect ablated data for resnet34
252fdea
@osw282
Collect ablated data for resnet34
33bbd5b
@d-lowl
Prepare running an experiment for resnet18 with fixed random seed
dcac052
@d-lowl
Prepare running an experiment for resnet18 with fixed random seed
6fdbfed
@osw282
Add partial resnet18 data
d67d69d
@osw282
Revert to the previous version of raw_data, fixing the seed does not …
e7d56fd 
…change it
@d-lowl
Updated notebook]
6667094
@d-lowl
Return zeros instead of reshaping
35106a3
@d-lowl
Add debug prints
2c73156
@osw282
Test data for updated ablation
ddb3c53
@d-lowl
Debug ablation
9f7ecc9
d-lowl and others added 25 commits January 9, 2025 16:36
@d-lowl
make zero tensor a parameter
86631d4
@d-lowl
make zero tensor a parameter
5336b17
@d-lowl
make zero tensor a parameter
d22c8ba
@d-lowl
forward random tensor
e9aa6f6
@d-lowl
forward random tensor
a6a6122
@d-lowl
make random ablation tensor with probing
a919957
@d-lowl
run benchmark
fc31744
@d-lowl
run benchmark
7194c05
@d-lowl
run benchmark
c827e95
@d-lowl
run benchmark
61d8056
@d-lowl
Enable benchmarking correctly
617c1de
@d-lowl
Enable benchmarking correctly
0665896
@osw282
debug prints
332665a
@d-lowl
Enable benchmarking correctly
6d05331
@d-lowl
Enable benchmarking correctly
4ecf994
@d-lowl
Remove debug prints
ca74184
@osw282
Remove raw test data
3dc258e
@d-lowl
Save engine correctly
94664ea
@d-lowl
Merge branch 'SD-94-ablation-study' of github.com:fuzzylabs/edge-visi…
de76f4f 
…on-power-estimation into SD-94-ablation-study
@d-lowl
Skip already done layers
51fd531
@d-lowl
Skip already done layers
f63c1b8
@d-lowl
Make ablated module depend on the input
23b1e6d
@osw282
Ablated results
2f4df7b
@d-lowl
Run measurements for shorter time, but longer warmup
f7dab14
@osw282
Ablated measurements
c4af926
@dagshub DagsHub
Copy link
Contributor

dagshub bot commented Jan 10, 2025

Join the discussion on DagsHub!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@d-lowl @osw282