Version 0.9.6: Notebooks / Model Performance Evaluation #10

josiahls · 2019-10-26T02:50:15Z

Version 0.9.0 is a near ready version of the repo. This will be benchmarking model performance, and showing basic repo use. One of the final major objects that are missing is the Interpretation object. The previous one was a bunch of spaghetti code fit for a high-end Italian restaurant. We need to simplify it for now. One of the most important things is allowing an interpretation object to merge with another interpretation object to compare reward graphs. The goal will be to do this in parallel with making the jupyter notebooks for this. We will also begin finalizing the README.

Goals:

Edit 10/27/2019:

Added Interpreter and GroupInterpreter for easy reward logging
Edit 12/14/2019:
README changes will be made for the 1.0 pull request since hopefully the code will be more consistent
Edit 12/15/2019:
Will push this to pypi I think so it is easier for people to download / install (?). Interested in testability.

- Interpreter (Cleaner) with cleaner code / closer to fastai - To and From Pickle - to_csv Notes: - I tried doing a from_csv implimentation, however I am seeing that something like this might not be possible unless using system system stuff. Not sure when I will ever get to this. I have some ideas about saving images / states as files with file paths... Maybe to_csv generates a file system also?

- Group Interpreter for combining model runs - Initial fixed dqn notebook (soft of) Fixed: - recorder callback ordering - renaming. It seems that fasti has some cool in-notebook test widgets that we might want to use in the future

- Group Interpreter merging - DQN base notebook - Interpreters with by default close envs Fixed: - env closing

- setup.py fastai needs to be min 1.0.59

- cpu / device issues.

- DQN Group Results - Reward Metric - Pipeline object - The idea is you have a fast_rl pipeline in a single function call, then the pipeline object just runs them on separate threads Notes: - I am realizing that we need sum reward smoothing. The graphs are way too messy.

- Analysis property to the group interpretation - Pipeline is demoable

- PER crashing due to containing 0 items

- Group Interpretation value smoothing

- Value smoothing making the reward values way too big

- Tests take too long. If Image input, just do a shorter fit cycle

- PER batch size not updating

- Tests take too long. If Image input, just do a shorter fit cycle

- cuda issues

- Tests take too long. If Image input, just do a shorter fit cycle

- Bounds n_possible_values is only calculated when used. Should make iteration faster. Added: - Smoothing for the scalar plotting

- cuda issues

- old code from README. Revisions coming.

- batch norm toggling. For now / forever defaulted to false

- revised test script

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

- Slowly adding tests.

- somehow trained_learner method in test was completely broken

- Slowly adding tests.

- Interpreter edge control. can also show average line

- models being all shitty. Apparently, batch norm reaaally screws them up. If you use batch norm, the batch size needs to be massive (128 wasnt large enough). By default, you can kind of turn off batch_norm in the Tabular models, but they still, when given a continuous input, will have an entry batch norm. I over-wrote it and now they work significantly better :)

- gitignore

Josiah Laivins added 6 commits October 25, 2019 22:37

Init Branch

906a47e

Added:

fc6af3c

- Group Interpreter for combining model runs - Initial fixed dqn notebook (soft of) Fixed: - recorder callback ordering - renaming. It seems that fasti has some cool in-notebook test widgets that we might want to use in the future

Added:

326328d

- Group Interpreter for combining model runs - Initial fixed dqn notebook (soft of) Fixed: - recorder callback ordering - renaming. It seems that fasti has some cool in-notebook test widgets that we might want to use in the future

Added:

694d76e

- Group Interpreter merging - DQN base notebook - Interpreters with by default close envs Fixed: - env closing

Added:

2c6dd98

- Group Interpreter merging - DQN base notebook - Interpreters with by default close envs Fixed: - env closing

josiahls added the enhancement New feature or request label Oct 28, 2019

josiahls and others added 23 commits October 28, 2019 12:40

Fixed:

9e2336c

- setup.py fastai needs to be min 1.0.59

Fixed:

7f77303

- cpu / device issues.

Added:

2ec8bb7

- DQN Group Results - Reward Metric - Pipeline object - The idea is you have a fast_rl pipeline in a single function call, then the pipeline object just runs them on separate threads Notes: - I am realizing that we need sum reward smoothing. The graphs are way too messy.

Added:

55a7ed5

- DQN Group Results - Reward Metric - Pipeline object - The idea is you have a fast_rl pipeline in a single function call, then the pipeline object just runs them on separate threads Notes: - I am realizing that we need sum reward smoothing. The graphs are way too messy.

Added:

331da03

- Analysis property to the group interpretation - Pipeline is demoable

Fixed:

54cfd20

- PER crashing due to containing 0 items

Added:

8400d9b

- Group Interpretation value smoothing

Merge remote-tracking branch 'origin/version_0_9_0' into version_0_9_0

7a62beb

Fixed:

ac88969

- Value smoothing making the reward values way too big

Fixed:

49f7ddb

- Tests take too long. If Image input, just do a shorter fit cycle

Fixed:

55c78e9

- PER batch size not updating

Merge remote-tracking branch 'origin/version_0_9_0' into version_0_9_0

15ca480

Fixed:

0edddf9

- Tests take too long. If Image input, just do a shorter fit cycle

Fixed:

78b1da2

- cuda issues

Fixed:

df0d729

- Tests take too long. If Image input, just do a shorter fit cycle

Merge remote-tracking branch 'origin/version_0_9_0' into version_0_9_0

801f3b2

Fixed:

5c1542c

- Bounds n_possible_values is only calculated when used. Should make iteration faster. Added: - Smoothing for the scalar plotting

More test fixing

d645716

More test fixing

3d3acff

Fixed:

5509344

- cuda issues

Merge remote-tracking branch 'origin/version_0_9_0' into version_0_9_0

d9582be

work

1a3e56c

work

f91f381

Josiah Laivins and others added 29 commits December 15, 2019 22:56

Removed:

2b0b265

- old code from README. Revisions coming.

Added:

afa2f4e

- batch norm toggling. For now / forever defaulted to false

Added:

a8fc416

- batch norm toggling. For now / forever defaulted to false

Added:

c3136a3

- batch norm toggling. For now / forever defaulted to false

Added:

9cb5014

- batch norm toggling. For now / forever defaulted to false

Added:

c33279b

- batch norm toggling. For now / forever defaulted to false

Added:

4a14965

- batch norm toggling. For now / forever defaulted to false

Version 0 9 5 mass refactor (#13)

0902506

Added:

11b210b

- revised test script

Added:

458dcf8

- revised test script

Added:

a280187

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

0e50fe6

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

e547ba4

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

7de393b

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

19f5be9

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

b3a6fb4

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

2f3cfd1

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

a84f5ad

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

5a747b5

- package caching. Reference https://docs.microsoft.com/en-us/azure/devops/pipelines/caching/?view=azure-devops. Not easy to understand since it is yarn (weird example right?)

Added:

151de34

- Slowly adding tests.

Fixed:

83c85dc

- somehow trained_learner method in test was completely broken

Merge remote-tracking branch 'origin/version_0_9_0' into version_0_9_0

19826f4

Added:

8563e1c

- Slowly adding tests.

Added:

412a831

- Slowly adding tests.

Added:

9d5d722

- Interpreter edge control. can also show average line

Updated:

1f2f497

- gitignore

Updated:

3b3a683

- gitignore

josiahls merged commit 0abb10a into master Dec 22, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Version 0.9.6: Notebooks / Model Performance Evaluation #10

Version 0.9.6: Notebooks / Model Performance Evaluation #10

Uh oh!

josiahls commented Oct 26, 2019 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Version 0.9.6: Notebooks / Model Performance Evaluation #10

Version 0.9.6: Notebooks / Model Performance Evaluation #10

Uh oh!

Conversation

josiahls commented Oct 26, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

josiahls commented Oct 26, 2019 •

edited

Loading