skills.rest
arthrod
arthrodCOMMUNITY
Explore all skills by this author →
GitHub

ml-gpu-training

Orchestrate GLiNER model training with hyperparameter optimization on AMD GPU droplets.

Updated Jul 18, 2026
Search Tags:#machine learning#fine-tuning#gpu#training#digitalocean#amd#gliner
One-click install
npx skills add https://github.com/arthrod/conejo-skills --skill ml-gpu-training
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ml-gpu-training
Source: https://github.com/arthrod/conejo-skills/tree/main/skills/ml-gpu-training
Command: npx skills add https://github.com/arthrod/conejo-skills --skill ml-gpu-training

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, flashdeberta, peft, python-dotenv, requests, wandb, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This skill solves the complexity of managing high-performance machine learning training workflows, specifically for GLiNER models, by automating hyperparameter strategy, remote infrastructure provisioning, and artifact management.

Core Features & Use Cases

  • Training Strategy: Provides a decision framework for hyperparameter tuning, LoRA vs. full fine-tuning, and loss debugging to ensure optimal model performance.
  • Remote-GPU Ops: Manages the lifecycle of AMD-based GPU droplets on DigitalOcean, including automated monitoring, artifact uploading to HuggingFace or S3, and server termination to prevent billing overruns.
  • Use Case: When fine-tuning a GLiNER model on a custom dataset, use this skill to validate your configuration, spin up an AMD MI300X droplet, monitor training progress via WandB, and automatically push the final checkpoint to HuggingFace before destroying the infrastructure.

Quick Start

Use the ml-gpu-training skill to validate your training configuration and launch a new GPU training run on a DigitalOcean droplet.

Frequently Asked Questions about ml-gpu-training

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fine-tune a GLiNER model using a remote AMD GPU on DigitalOcean?▼

Fine-tuning a GLiNER model on a remote AMD GPU is handled by automating DigitalOcean droplet provisioning, executing the training lifecycle, and uploading artifacts to HuggingFace before terminating the server.

What is the best way to manage remote GPU training infrastructure and prevent billing overruns?▼

Managing remote GPU training infrastructure requires monitoring the lifecycle of AMD GPU droplets, tracking progress via WandB, and automatically terminating servers after artifact persistence to prevent billing overruns.

Does this training workflow support both LoRA and full fine-tuning for GLiNER models?▼

Yes, the training workflow supports both approaches by providing a decision framework to evaluate hyperparameter tuning strategies and choose between LoRA and full fine-tuning for optimal model performance.

How do I automatically push trained model checkpoints to HuggingFace after a GPU training run?▼

Automatically pushing trained model checkpoints to HuggingFace is integrated into the remote GPU ops workflow, triggering artifact uploads upon training completion before the infrastructure is destroyed.

Can I use WandB to monitor hyperparameter optimization during deep learning training?▼

Yes, you can use WandB to monitor hyperparameter optimization and track training progress seamlessly, ensuring rigorous validation throughout the deep learning research and production fine-tuning process.

Why do I need automated server termination after pushing training artifacts to S3 or HuggingFace?▼

Automated server termination is needed to enforce cost-efficient resource lifecycle management, immediately destroying high-memory GPU compute instances once artifact uploading to HuggingFace or S3 completes.

Related Skills

Explore More Software Engineering Skills→
maziyarpanahimaziyarpanahi
5.0k

running-zeroshot-ner

Extract custom biomedical entities from text using local GLiNER checkpoints.

Community
Intermediate
ScientiaCapitalScientiaCapital
1

model-deployment

Export fine-tuned models to GGUF and deploy with vLLM or Ollama.

Community
Advanced
Ced3-hanCed3-han

peft-fine-tuning

Fine-tune large language models with LoRA and QLoRA on limited GPU memory.

Community
Advanced
bobmatnycbobmatnyc
68

digitalocean-compute

Manage DigitalOcean compute services including Droplets, Kubernetes, and GPU instances.

Community
Intermediate