shaikhsalman/devsecops-platform

1---2license: apache-2.03tags:4- devsecops5- llm6- sft7- lora8- tulu-39- kubernetes10- terraform11---12 13# DevSecOps Model Platform14 15> Train a secure model on the best data, then deploy it securely.16 17## Start Here: Train Your Model18 19| Dataset | Size | What It Gives You | Command |20|---------|------|-------------------|---------|21| **tulu-3-sft-mixture** | 940K | Math, code, safety, chat (BEST) | python model/train_tulu3.py |22| **OpenThoughts-114k** | 114K | Reasoning, chain-of-thought | python model/train_openthoughts.py |23 24**allenai/tulu-3-sft-mixture** is from Allen AI Tulu 3 - current SOTA open instruction-tuned model. Proven on Llama-3.1-8B: MMLU 53.5, GSM8K 79.9, HumanEval 76.8.25 26LoRA config from LoRA Without Regret (Schulman 2025): r=256, alpha=16, all-linear = matches full fine-tuning at 67% compute.27 28## Repository Structure29 30```31model/                     THE MODEL - train, serve, enhance32  train_tulu3.py             Primary: 940K best data (zero preprocessing)33  train_openthoughts.py      Reasoning: 114K CoT traces34  finetune_configurable.py   Multi-dataset configurable trainer35  rag_pipeline.py             RAG for DevSecOps knowledge36  DATASETS.md                 Why these datasets, proven recipes37 38deployment/               SERVE IT - Kubernetes + Docker + vLLM39  deployment.yaml             ML inference K8s manifest40  mlflow-deployment.yaml      Experiment tracking41  Dockerfile.ml-inference     Hardened multi-stage image42 43security/                 PROTECT IT - scanning + policies44  scanning/                   Trivy, Semgrep, Checkov, SBOM45  policies/                   Kyverno, OPA Gatekeeper46 47infrastructure/           RUN IT - Terraform + monitoring + CI/CD48  terraform/                  VPC, EKS, RDS, S3, IAM, KMS, GuardDuty, Macie49  monitoring/                 Prometheus, Alertmanager, OTEL, Grafana50  ci-cd/                      GitHub Actions DevSecOps pipeline51 52compliance/               CERTIFY IT - SOC2, NIST, CIS53  controls-mapping.yaml       SOC2 Type II54  nist-800-53-mapping.yaml    NIST 800-53 Rev555  cis-eks-k8s.yaml            CIS Benchmarks56```57 58## Quick Commands59 60```bash61# Train on best data (A100, ~6h)62python model/train_tulu3.py63 64# Quick test (any GPU)65python model/train_tulu3.py --max_steps 100 --no_push66 67# Security scan68python security/scanning/security_audit.py69 70# Deploy model to K8s71kubectl apply -f deployment/deployment.yaml72 73# Infrastructure (Terraform)74cd infrastructure/terraform/environments/prod && terraform apply75```76