Fix the accuracy issue by update celu formula based on spec

[sgbihu]

Details:

• The implementation didn't align with ONNX spec.

max(0,x) + min(0,alpha*(exp(x/alpha)-1))

Tickets:

• CVS-178958

[Copilot]

Pull request overview

This PR fixes the CELU (Continuously Differentiable Exponential Linear Unit) implementation to match the ONNX specification formula: max(0,x) + min(0,alpha*(exp(x/alpha)-1)). The previous implementation incorrectly used the ELU operator instead of explicitly implementing the CELU formula.

Key Changes:

• Replaced ELU-based implementation with explicit formula computation using maximum, minimum, exponential, and arithmetic operations
• Added necessary OpenVINO operation headers (Add, Exp, Maximum, Minimum, Subtract)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rkazants]

please add layer tests reproducing this issue

[mvafin]

I see celu also exist in pytorch. Can you check if the same fix is needed for pytorch?

[sgbihu]

I see celu also exist in pytorch. Can you check if the same fix is needed for pytorch?

Fixed.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.