/llmops-conference-session

Presentation, Code and Notebooks used in the conference

Primary LanguageJupyter Notebook

Notebooks and Code

Sesion : LLMOps – The Next Frontier of Scaling Generative AI Powered Applications

Speaker : Abhishek Kumar

Notebooks

  1. Langchain Prompt Tracking with LangSmith and Weights & Biases
  2. Download model weights from HuggingFace and Pushing to GCS
  3. Use hosted LLM API with Langchain
  4. Fine tuning LLAMA-2 models
  5. Using Argilla for Data Preperation ( for RLHF, Fine-Tuning)
  6. 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