Help Center - EzEpoch | ARC Technologies LLC

📸 Platform Screenshots — See What You're Getting

A complete visual walkthrough of every screen in EzEpoch

Login & Account

🔐 Login Page Secure sign-in with email/password. No credit card needed for the free trial.

Training Configuration

⚙️ Default View Pick a model and GPU — smart defaults fill everything else in automatically.

🤖 Model Selected Text model selected with GPU star system showing which GPUs can fully train it.

🔬 Advanced Settings Precision, optimizer, Flash Attention — incompatible options auto-greyed out for your GPU.

📖 All Settings Complete view with API keys, optimizer table, and all configuration panels expanded.

🧪 Expert Settings Multi-GPU strategy (DDP/FSDP/ZeRO), LoRA rank tuning, warmup steps, and advanced parameters.

Requirements & Package Builder

📋 Requirements Auto-generated dependency list with PyTorch/CUDA compatibility presets and conflict checks.

📦 Build Package One-click training package creation — validated and conflict-free, ready to deploy anywhere.

Cloud Deploy

☁️ Cloud Deploy Browse live GPUs from RunPod and Vast.ai with real-time pricing — deploy with one click.

🔑 SSH & Providers Custom SSH support for AWS, GCP, Lambda Labs, or any server you control.

🚀 Deploy Status Real-time deployment progress — environment setup, file transfer, and training launch.

Live Dashboard & AI Monitoring

📊 Dashboard Overview AI Guardian alerts, training status, live metrics, and session management in one view.

📈 Live Charts Training/validation loss, GPU utilization, memory usage, and learning rate curves updated in real-time.

🔍 Advanced Metrics Gradient norms, training speed, ETA estimates, and AI Guardian insight overlays.

EzSetup — Dependency Resolver

🔧 EzSetup Paste your requirements.txt — auto-detect conflicts, fix versions, and generate a correct install script.

🧙 Setup Wizard - Step-by-Step Guide (9 Pages)

The wizard walks you through everything you need, one step at a time

The Setup Wizard is perfect for beginners or when you want guided configuration. It breaks down the process into 9 easy pages.

1

Welcome

Choose to create a New Training Package or use Quick Deploy with an existing package.

2

API Configuration

Enter all your API keys (HuggingFace, OpenAI, etc.) and generate an SSH key for cloud deployment. We include an SSH key generator to make this easy!

3

Model Type

Select your training type: Text, Vision, Audio, or Real Multimodal. Custom configurations are available in the full App.

4

Model Selection

Choose your actual model like "Falcon 7B", "Llama 3 8B", "Mistral 7B", etc. The system shows memory requirements and compatibility.

5

GPU Configuration

Select the GPU you plan to use (RTX 4090, A100, H100, etc.). The system calculates optimal batch size and memory settings automatically.

6

Data Processing Mode

Choose how to handle your training data: Chunking, Progressive Examples, streaming options, and more.

7

Requirements Generation

Automatically generates all Python dependencies (66+ packages) with conflict checking and security analysis.

8

Training Package Generation

Creates your complete training package with all scripts, configurations, and the AI Guardian monitoring system.

9

Select Cloud GPU

Browse available GPUs on RunPod, Vast.ai, or use Custom SSH. Once your instance is running, click on it to start the Deployment Wizard.

💡 Pro Tip: After selecting your cloud GPU instance and it's running, click on it to launch the Deployment Wizard. This same deployment flow works for the Wizard, App, and Quick Deploy!

⚖️ App vs Wizard - Which Should You Use?

Understanding the difference between the full App and the Setup Wizard

Feature	🧙 Setup Wizard	🖥️ Full App
Best For	Beginners, quick setups	Power users, custom configs
Pages/Steps	9 guided pages	Multiple tabs, all at once
Advanced Settings	Simplified defaults	Full access to all options
Custom Multimodal	Preset options only	Full custom configuration
GPU Calculations	✅ Same algorithms	✅ Same algorithms
AI Guardian	✅ Included	✅ Included
Deployment Flow	✅ Same deployment wizard	✅ Same deployment wizard

🎯 Our Recommendation: Start with the Setup Wizard for your first training run. Once you're comfortable, the Full App gives you access to all advanced options like custom multi-GPU strategies (DDP, FSDP, ZeRO), precision settings, and more.

🚀 Deployment Flow - From Config to Training

How to deploy your training package to cloud GPUs

The Correct Order:

Build Config - Use Wizard or App to configure your training
Select Cloud GPU - Browse and rent a GPU from RunPod, Vast.ai, or Custom SSH
Choose Training Data - Select or upload your dataset
Deploy - Click deploy and watch AI Guardian take over!

⚠️ Important: You do NOT download ZIP files anymore! EzEpoch deploys directly to your cloud GPU via SSH. Your training data stays on your local machine or cloud storage - we never store it on our servers.

What Happens During Deployment:

1. Connection: EzEpoch connects to your GPU instance via SSH
2. Environment: Sets up Python, CUDA, and all dependencies automatically
3. Data Transfer: Securely streams your training data (never stored on our servers)
4. Training Start: Launches training with AI Guardian monitoring active
5. Real-Time Monitoring: Watch progress in your Dashboard with AI insights

🛡️ AI Guardian - Patent Pending Technology

Revolutionary AI monitoring that predicts and prevents training failures

AI Guardian is EzEpoch's patent-pending technology that uses real AI analysis to monitor your training runs. Unlike other platforms that just show metrics, we actively predict, prevent, and recover from issues.

🚀 What AI Guardian Does:

🔮 Crash Prediction (Patent Pending): Analyzes GPU memory, gradients, and training patterns to predict failures BEFORE they happen
🛡️ Auto-Recovery (Patent Pending): If a crash does occur, automatically recovers from the last checkpoint without losing progress
📊 Real-Time Analysis: Provides intelligent insights about your training performance, not just raw numbers
💡 Smart Recommendations: Suggests parameter adjustments to improve training speed and quality
📱 Mobile Monitoring: Check on your training from your phone - get alerts and control remotely

🏆 99% Training Success Rate!
With AI Guardian, training failures drop from 40% (industry average) to less than 1%. That's real AI doing real work to protect your time and money.

☁️ Cloud GPU Setup Guide

How to connect RunPod, Vast.ai, or Custom SSH servers

Supported Providers:

🟣 RunPod

Enter your RunPod API key. EzEpoch will list available GPUs, pricing, and let you deploy with one click.

🔵 Vast.ai

Enter your Vast.ai API key. Browse the GPU marketplace directly within EzEpoch.

🟡 Custom SSH

Use any server with SSH access - AWS, GCP, your own hardware, or any other provider.

💰 GPU Costs Are Separate: EzEpoch subscription covers the platform, AI Guardian, and tools. You pay GPU providers (RunPod, Vast.ai, etc.) directly for compute time. We don't markup GPU costs!

🔧 Troubleshooting Common Issues

Quick fixes for the most common problems

SSH Connection Failed: Make sure your SSH key is added to the cloud provider. Use our built-in SSH key generator in the wizard.
Out of Memory (OOM): AI Guardian should prevent this, but if it happens, try: smaller batch size, enable gradient checkpointing, or use a larger GPU.
Training Stuck: Check the Dashboard for AI Guardian alerts. It may be waiting for data or encountering a known issue.
API Key Invalid: Double-check your HuggingFace/OpenAI keys. Some models require accepting license agreements on HuggingFace first.
Model Not Found: Ensure the model name matches exactly (case-sensitive). Check if it's a gated model requiring special access.

📧 Still stuck? Contact the developer directly at wil@ezepoch.com - as a solo developer, I personally respond to every support request!

📚 Help Center