feature:verl

[yuehua-s]

No description provided.

[Copilot]

Pull Request Overview

This PR introduces a new Helm chart for deploying VERL (Volcano Engine Reinforcement Learning) on TKE (Tencent Kubernetes Engine) and updates an existing LLaMA-Factory chart configuration. The changes create a complete Kubernetes deployment structure for VERL with GPU support and shared memory configuration.

• Adds a complete VERL-on-TKE Helm chart with deployment, service, ingress, and autoscaling templates
• Updates LLaMA-Factory chart to use Tencent mirrors and improved resource allocation
• Configures both charts with optimized GPU resource requests and shared memory volumes

Reviewed Changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
incubator/verl-on-tke/values.yaml	Configuration values for VERL deployment with GPU resources and environment variables
incubator/verl-on-tke/templates/serviceaccount.yaml	ServiceAccount template for VERL deployment
incubator/verl-on-tke/templates/service.yaml	Service template exposing VERL application
incubator/verl-on-tke/templates/ingress.yaml	Ingress template for external access to VERL
incubator/verl-on-tke/templates/hpa.yaml	Horizontal Pod Autoscaler template for VERL scaling
incubator/verl-on-tke/templates/deployment.yaml	Main deployment template for VERL containers
incubator/verl-on-tke/templates/_helpers.tpl	Helm template helpers for naming and labeling
incubator/verl-on-tke/templates/NOTES.txt	Post-installation notes for accessing the VERL application
incubator/verl-on-tke/Chart.yaml	Chart metadata and version information for VERL
incubator/verl-on-tke/.helmignore	File patterns to ignore during chart packaging
incubator/llama-factory-on-tke/values.yaml	Updated resource limits and PIP index configuration
incubator/llama-factory-on-tke/Chart.yaml	Improved chart description and removed unused sources

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The containerPort is using the service port value, but these should be independent. The container should expose its own port (typically 8080 or 3000), and the service should map to that container port. This creates a tight coupling between service and container configuration.

Suggested change

	containerPort: {{ .Values.service.port }}
	containerPort: {{ .Values.containerPort }}

[Copilot]

The current configuration runs an infinite sleep loop which prevents the container from serving actual traffic. This setup appears to be for development/debugging purposes but may not be suitable for production deployments.

Suggested change

	args:
	- trap 'exit 0' TERM; sleep infinity & wait
	# args:
	# - <your-app-command> # Specify the actual application command here if needed

[pavlelee]

lgtm