Fixing cpu-or-gpu parameter bug to enable gaudi3

[toddtomashek-c2l]

The input value of "gaudi3" is not preserved in the "gaudi_platform" variable because the "cpu_or_gpu" value is reset to "g" after initial processing. Subsequent calls of read_config_file() will see 'cpu_or_gpu=="g"' and reset gaudi_platform to "gaudi2".

This PR prevents that by only setting "gaudi_platform" the first time it is called.

[mdfaheem-intel]

We are currently in process to review this fix, could you please share the detailed steps to reproduce this issue at my end.

[toddtomashek-c2l]

You can reproduce by using the menu to deploy models (we saw using deepseek-r1-qwen) when the inference-config file does not have ONLY "deploy_llm_models" = "on." For example, on execution of the script for an initial deployment that includes "deploy_kubernetes_fresh=on", "deploy_ingress_controller=on" and "deploy_llm_models=on" with a "cpu-or-gpu" value of "gaudi3", the "read_config_file()" function will be called multiple times and the "gaudi3" values will be overwritten to "gaudi2" because after the first execution of "read_config_file()" the cpu_or_gpu variable will be changed from "gaudi3" (original input) to "g"

[mdfaheem-intel]

You can reproduce by using the menu to deploy models (we saw using deepseek-r1-qwen) when the inference-config file does not have ONLY "deploy_llm_models" = "on." For example, on execution of the script for an initial deployment that includes "deploy_kubernetes_fresh=on", "deploy_ingress_controller=on" and "deploy_llm_models=on" with a "cpu-or-gpu" value of "gaudi3", the "read_config_file()" function will be called multiple times and the "gaudi3" values will be overwritten to "gaudi2" because after the first execution of "read_config_file()" the cpu_or_gpu variable will be changed from "gaudi3" (original input) to "g"

Thanks for sharing the detailed steps, let me validate this fix and provide you the update