Update: fnn full cov unit test results

[zhanwen-xin]

fnn full cov results for omega_tol = 0.001f.

[lhnguyen102]

@zhanwen-xin and @jamesgoulet, I have to reverted the PR #36 because the test didn't pass because the unittest for full covariance are generated with the original omega_tol. In addition, I have added an additional requirements where all the test must be passed before merging to main. Here is the instructions to locally run the test to ensure all tests are passed. For the instability issue, @zhanwen-xin I would suggest treating omega_tol as user-specified input where it can be modified in the config file.

layers:     [1,     1,       1,     1]
nodes:      [1,     30,      30,    1]
activations:    [0,     7,       7,     0]
...
omega_tol: 0.001

Here is the function to be modified. See how I did with other variables e.g., sigma_x and do the same thing to omega_tol. Here is PR template for any PR.

Description

What is this PR for?

Changes Made

What changes have been done?

Note for Reviewers

How I can run the test for your changes if applied

[jamesgoulet]

Ha,

Thank you for your help. We tried to fixed it today and reached out to Miquel but you fixed the issue first. 🙏 We will get better over time…

Thank you again!

James

On Sep 15, 2023, at 17:57, Luong-Ha Nguyen @.***> wrote:

@zhanwen-xin https://github.com/zhanwen-xin and @jamesgoulet https://github.com/jamesgoulet, I have to reverted the PR #36 https://github.com/lhnguyen102/cuTAGI/pull/36 because the test didn't pass because the unittest for full covariance are generated with the original omega_tol. In addition, I have added an additional requirements where all the test must be passed before merging to main. Here is the instructions https://github.com/lhnguyen102/cuTAGI/tree/main/test to locally run the test to ensure all tests are passed. For the instability issue, @zhanwen-xin https://github.com/zhanwen-xin I would suggest treating omega_tol as user-specified vinput where it can be modified in the config file https://github.com/lhnguyen102/cuTAGI/blob/main/cfg/1fc_full_cov_toy.txt.

layers: [1, 1, 1, 1] nodes: [1, 30, 30, 1] activations: [0, 7, 7, 0] ... omega_tol: 0.001 Here is the function https://github.com/lhnguyen102/cuTAGI/blob/main/src/net_prop.cpp#L1211 to be modified. See how I did with other variables https://github.com/lhnguyen102/cuTAGI/blob/main/src/net_prop.cpp#L1307 e.g., sigma_x and do the same thing to omega_tol. Here is PR template for any PR.

Description

What is this PR for?

Changes Made

What changes have been done?

Note for Reviewers

How I can run the test for your changes if applied

— Reply to this email directly, view it on GitHub https://github.com/lhnguyen102/cuTAGI/pull/37#issuecomment-1721903438, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGG5HOP4NDZVJUWVLFCPDWLX2TFNLANCNFSM6AAAAAA42DEEYY. You are receiving this because you were mentioned.

[lhnguyen102]

James, no worries! No one can't run the code without bugs :)

[zhanwen-xin]

Ha,

I will try that. Thanks a lot!

Zhanwen

On Sep 15, 2023, at 6:02 PM, James-A. Goulet @.***> wrote:

Ha,

Thank you for your help. We tried to fixed it today and reached out to Miquel but you fixed the issue first. 🙏 We will get better over time…

Thank you again!

James

On Sep 15, 2023, at 17:57, Luong-Ha Nguyen @.***> wrote:

@zhanwen-xin https://github.com/zhanwen-xin and @jamesgoulet https://github.com/jamesgoulet, I have to reverted the PR #36 https://github.com/lhnguyen102/cuTAGI/pull/36 because the test didn't pass because the unittest for full covariance are generated with the original omega_tol. In addition, I have added an additional requirements where all the test must be passed before merging to main. Here is the instructions https://github.com/lhnguyen102/cuTAGI/tree/main/test to locally run the test to ensure all tests are passed. For the instability issue, @zhanwen-xin https://github.com/zhanwen-xin I would suggest treating omega_tol as user-specified vinput where it can be modified in the config file https://github.com/lhnguyen102/cuTAGI/blob/main/cfg/1fc_full_cov_toy.txt.

layers: [1, 1, 1, 1] nodes: [1, 30, 30, 1] activations: [0, 7, 7, 0] ... omega_tol: 0.001 Here is the function https://github.com/lhnguyen102/cuTAGI/blob/main/src/net_prop.cpp#L1211 to be modified. See how I did with other variables https://github.com/lhnguyen102/cuTAGI/blob/main/src/net_prop.cpp#L1307 e.g., sigma_x and do the same thing to omega_tol. Here is PR template for any PR.

Description

What is this PR for?

Changes Made

What changes have been done?

Note for Reviewers

How I can run the test for your changes if applied

— Reply to this email directly, view it on GitHub https://github.com/lhnguyen102/cuTAGI/pull/37#issuecomment-1721903438, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGG5HOP4NDZVJUWVLFCPDWLX2TFNLANCNFSM6AAAAAA42DEEYY. You are receiving this because you were mentioned.

— Reply to this email directly, view it on GitHub https://github.com/lhnguyen102/cuTAGI/pull/37#issuecomment-1721910343, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYCKBXC7O5UFYZW7LZFGIDLX2TGBBANCNFSM6AAAAAA42DEEYY. You are receiving this because you were mentioned.

[lhnguyen102]

You're welcome! ping me if needed