There are layers, there are protocols, and there are mechanisms.
GBN, SR, etc are examples of mechanisms. These can be implemented by a protocol to achieve reliability. There are different protocols that use different mechanisms.
Protocols usually are assigned to layers. There are layer 2, layer 3, and layer 4 protocols.
Each of the layers can employ a protocol that uses mechanisms such as acknowledgements and retransmissions. Protocols on different layers are (in theory) independent of each other. Actually, protocols are usually just independent of each other, unless they are specifically designed to work together and together only. If several protocols use acknowledgements, these acknowledgements are independent of each other. Layer 2 will employ a layer 2 protocol that has its own acknowledgments. These acknowledgments are completely independent of any other acknowledgements of any other protocols that any other layer may or may not use. For layer 2, TCP acks are just packets, nothing special. They only have meaning for the TCP instance on the other side.
I followed from this question understand that only transport layer responsible for acknowledgement.
No. First, OSI model is designed to reason about networks in general, not Internet in particular. A network can have reliable protocols on any layer, including layer 3.
Internet is a particular network, where layer 3 is not reliable, and does not use any mechanism. In Internet, reliable end-to-end delivery needs to be achieved by layer 4 or above. This is because layer 3 does not provide any guarantees even if a specific layer 2 does.
Layer 2 can provide relible delivery between two layer 2 hops. This is reliable delivery on a single layer 2 hop, not end-to-end. The packet can still be lost on any other hop along the way. This is why we need a separate mechanism fo reliable end-to-end delivery.