Normalize input tensor names

[nnshah1]

updating to make names lower case and change 'prompt' and 'text' to more generic:

'text_input', 'text_output'

[rmccorm4]

LGTM, but need to sync with @nv-hwoo @matthewkotila for PA VLLM guide updates, and @pskiran1 for VLLM CI test updates.

L0_http/generate_endpoint_test.py uses a mock model so it can be updated separately, should be unaffected by this.

[matthewkotila]

@rmccorm4: LGTM, but need to sync with @nv-hwoo @matthewkotila for PA VLLM guide updates, and @pskiran1 for VLLM CI test updates.

L0_http/generate_endpoint_test.py uses a mock model so it can be updated separately, should be unaffected by this.

triton-inference-server/client#412 ready to go 👍🙏

[nv-hwoo]

One nit comment but LGTM. I ran the changes with PA LLM guide and confirmed that it works 👍

[nnshah1]

@tanmayv25, @jbkyang-nvi let me know if we want to merge this or close it as the changes have merged into the backend.

[matthewkotila]

What's still blocking this?

[matthewkotila]

@nnshah1 @tanmayv25 @rmccorm4 @jbkyang-nvi Friendly ping on what is blocking this?

[nnshah1]

@nnshah1 @tanmayv25 @rmccorm4 @jbkyang-nvi Friendly ping on what is blocking this?

Plan is to move this tutorial to reference the new backend and the changes for naming have already been made there. We wanted to stop making changes here to avoid duplication / sync issues.

@matthewkotila Let us know issues with that approach.