Notebooks
- Langchain Prompt Tracking with LangSmith and Weights & Biases
- Download model weights from HuggingFace and Pushing to GCS
- Use hosted LLM API with Langchain
- Fine tuning LLAMA-2 models
- Using Argilla for Data Preperation ( for RLHF, Fine-Tuning)
- Creating Programmatic Guardrails for LLMs
Note : replace the keys and tokens to excute the python codes and YAML files.
Demos were tested in GKE Autopilot. Once the Kubernetes cluster is configured.
# setup storage class
kubectl apply -f deployment/storage.yaml
# deploy llama 2 model
kubectl apply -f deployment/llama2-v1-deployment.yaml
# deploy fine-tuned llama 2 model
kubectl apply -f deployment/llama2-v2-deployment.yaml
# deploy argilla
kubectl apply -f deployment/argilla.yaml