overbrainer 0.8.2

Distill a large LLM into a smaller open-weights model: generate questions, collect answers, fine-tune, visualize.
Documentation
1
2
3
4
5
6
7
8
9
10
11
# Copy to .env (gitignored) or export in your shell.
# Values may be literals or Vault references: vault:<mount>/<path>#<field>
OVERBRAINER_PROVIDERS__OPENROUTER__BASE_URL=https://openrouter.ai/api/v1
OVERBRAINER_PROVIDERS__OPENROUTER__API_KEY=vault:secret/overbrainer/openrouter#api_key
# OVERBRAINER_TARGETS__HOMELAB__HOST=user@gpu-box
# OVERBRAINER_RUNPOD__API_KEY=vault:secret/overbrainer/runpod#api_key
# OVERBRAINER_TARGETS__GPU_CLOUD__GPU_TYPES=NVIDIA GeForce RTX 4090,NVIDIA A40
# OVERBRAINER_HF_TOKEN=          # gated or private base models, and training.hub_model_id
# OVERBRAINER_LOG=info
# VAULT_ADDR=https://vault.example.com
# VAULT_TOKEN=