mlops-suite

MLOps tools — model serving (vLLM), local inference (llama.cpp/GGUF), model hub (HuggingFace), experiment tracking (Weights & Biases), audio generation (AudioCraft), image segmentation (SAM), and LLM

SKILL.md

GitHub

MLOps Suite

Covers: model serving, local inference, model hub, experiment tracking, audio/image generation.

Model Serving (vLLM)


set -euo pipefail
# High-throughput LLM serving with OpenAI API
python -m vllm.entrypoints.openai.api_server \
  --model meta-llama/Llama-3.1-8B --dtype auto

Local Inference (llama.cpp / GGUF)


set -euo pipefail
# GGUF inference
llama-cli -m model.gguf -p "Hello" -n 128

HuggingFace Hub


set -euo pipefail
hf search "llama"
hf download meta-llama/Llama-3.1-8B
hf upload ./my-model

Experiment Tracking (Weights & Biases)

import wandb
wandb.init(project="my-project")
wandb.log({"loss": 0.5})

set -euo pipefail

Audio Generation (AudioCraft / MusicGen)

# Text-to-music, text-to-sound
# See references/audiocraft.md

set -euo pipefail

Image Segmentation (SAM)

# Zero-shot segmentation via points, boxes, masks
# See references/sam.md

LLM Evaluation

See local-ml-hosting skill for lm-eval-harness details.