Add Qwen2 example to the trainer

[jaiakash]

What you would like to be added?

We need an additional example for training Qwen3 models in TorchTune that does not require a Hugging Face token.

Recipes: https://github.com/pytorch/torchtune/blob/main/recipes/configs/qwen2/1.5B_full.yaml
HF: https://huggingface.co/Qwen/Qwen2-1.5B

Why is this needed?

• As part of kubeflow/trainer#2762, we want to run GPU workloads on OCI VMs. The current TorchTune examples require a Hugging Face token, which prevents running CI actions for each PR due to security concerns.
• An example that does not require a Hugging Face token allows CI testing for GPU workloads safely.
• Adds another reference example to the trainer repository for users.

Love this feature?

Give it a 👍 We prioritize the features with most 👍

[jaiakash]

/remove-label lifecycle/needs-triage
/assign @jaiakash
/area llm
/area examples
/area runtimes

[jaiakash]

Changed to qwen2 model