Ready-made Docker image for running coding agents on ML/fine-tuning jobs

Melvin Vivas · X post · 2026-09-02 · Open on X

Topics: Fine-tuning & Model Customization, AI Dev Tools & Productivity, LLMOps, Deployment & Monitoring · Level: intermediate

Summary

For ML work like fine-tuning with coding agents, the creator offers a ready-made Docker image with PyTorch/CUDA and the agents already installed. You run it with GPU access, log in to the agents, and start working. It also includes herdr for persistent sessions.

Key points

• Run: docker run --rm -it --gpus all melvindave/runpod-pytorch:1.1.0
• --gpus all exposes your NVIDIA GPUs to the container; --rm removes the container when it exits
• Nothing to install after starting it: just log in to the coding agents
• herdr is included to keep sessions running
• Meant for ML tasks like fine-tuning done with coding agents

Resources mentioned

• melvindave/runpod-pytorch Docker image · tool · hub.docker.com · free
  Docker image for GPU ML development with coding agents, the Hugging Face CLI and nvtop pre-installed.
  Also in: GPU devbox Docker image with coding agents on Runpod (Melvin Vivas on X · notes), Reusable GPU devbox: PyTorch/CUDA plus six coding agents (Melvin Vivas on X · notes)
• herdr · tool · x.com · check price
  A runtime that keeps real terminals open for coding agents on a local or rented machine, so agent sessions keep running when you disconnect.
  Also in: Terminal Tool Recommendation: herdr (Melvin Vivas on X · notes), herdr 0.9.0: Control Agents Across Multiple Machines (Melvin Vivas on X · notes), GPU devbox Docker image with coding agents on Runpod (Melvin Vivas on X · notes), GPU-Ready AI Devbox Docker Image on Runpod (PyTorch 2.8 + CUDA 12.8) (Melvin Vivas on X · notes) and 7 more

Try this

• Run melvindave/runpod-pytorch:1.1.0 with --gpus all when using coding agents for ML tasks like fine-tuning
• Log in to your coding agents inside the container
• Use a coding agent inside the GPU container to run a fine-tuning job