[CI] Set read-all for workflows top-level token-permissions

[mengfei25]

disable_all

[mengfei25]

Test workflow https://github.com/intel/torch-xpu-ops/actions/runs/15866174677

[dvrogozh]

@mengfei25 : this PR breaks accelerate and transformers tests scheduled to execute weekly:

• https://github.com/intel/torch-xpu-ops/actions/runs/16089832250

+ gh --repo intel/torch-xpu-ops pr view
+ tee pr-info.txt
argument required when using the --repo flag

Usage:  gh pr view [<number> | <url> | <branch>] [flags]

Flags:
  -c, --comments          View pull request comments
  -q, --jq expression     Filter JSON output using a jq expression
      --json fields       Output JSON with the specified fields
  -t, --template string   Format JSON output using a Go template; see "gh help formatting"
  -w, --web               Open a pull request in the browser
  
Error: Process completed with exit code 1.

The change you've introduced adds some mechanism to skip these tests, right? It does so based on commit/PR description? Unfortunately there is no any description provided to understand how it works. As such it's useless as can be used by limited number of people who knows how it works. Secondly, as I noted, it effectively breaks CI as it does not account on the scheduled workflows and handles only PRs.

Further, the 1777 PR title says to add read-all permissions. Why does it adjust any workflows to add enable/disable mechanism? That's the change unrelated to the subject. Even if some workflows are long/failing/etc. for permissions unrelated reasons - we should not go and start enabling/disabling them.

I think the right strategy in your case was to request an exception to merge the change on some failing cases in Accelerate/Transformers. You also could break that into 2 steps: modify all workflows except Accelerate/Transformers and then continue working (with me for example) to change the remaining 2.

Overall, I strongly suggest that we need to revert some of the changes from PR 1777 for accelerate and transformers. Once done, you can reintroduce the fixed and described changes. Please, review:

• ci: revert breaking changes in accelerate and transformers tests #1816

[mengfei25]

@mengfei25 : this PR breaks accelerate and transformers tests scheduled to execute weekly:

• https://github.com/intel/torch-xpu-ops/actions/runs/16089832250

+ gh --repo intel/torch-xpu-ops pr view
+ tee pr-info.txt
argument required when using the --repo flag

Usage:  gh pr view [<number> | <url> | <branch>] [flags]

Flags:
  -c, --comments          View pull request comments
  -q, --jq expression     Filter JSON output using a jq expression
      --json fields       Output JSON with the specified fields
  -t, --template string   Format JSON output using a Go template; see "gh help formatting"
  -w, --web               Open a pull request in the browser
  
Error: Process completed with exit code 1.

The change you've introduced adds some mechanism to skip these tests, right? It does so based on commit/PR description? Unfortunately there is no any description provided to understand how it works. As such it's useless as can be used by limited number of people who knows how it works. Secondly, as I noted, it effectively breaks CI as it does not account on the scheduled workflows and handles only PRs.

Further, the 1777 PR title says to add read-all permissions. Why does it adjust any workflows to add enable/disable mechanism? That's the change unrelated to the subject. Even if some workflows are long/failing/etc. for permissions unrelated reasons - we should not go and start enabling/disabling them.

I think the right strategy in your case was to request an exception to merge the change on some failing cases in Accelerate/Transformers. You also could break that into 2 steps: modify all workflows except Accelerate/Transformers and then continue working (with me for example) to change the remaining 2.

Overall, I strongly suggest that we need to revert some of the changes from PR 1777 for accelerate and transformers. Once done, you can reintroduce the fixed and described changes. Please, review:

• ci: revert breaking changes in accelerate and transformers tests #1816

Fixed in #1846
https://github.com/intel/torch-xpu-ops/pull/1846/files#diff-8fa2c037d3633c466cf7c27fff1d25ff32b8993a024a3b359166d720b2da815e