Skip to content

SPACES_GUIDE.md

Latest commit

 

History

History
131 lines (81 loc) · 2.96 KB

SPACES_GUIDE.md

File metadata and controls

131 lines (81 loc) · 2.96 KB

HuggingFace Spaces Deployment Guide

Overview

Deploy your sentiment analysis model as an interactive demo on HuggingFace Spaces using Gradio.

Live Demo: HuggingFace Spaces

Required Files

These files must be in your Space's root directory:

  1. app.py - Gradio application
  2. requirements.txt - All dependencies
  3. README.md - Space description (optional)

Quick Start

Option 1: GitHub Integration (Recommended)

  1. Fork this repository
  2. Create a new Space on HuggingFace
  3. Connect your GitHub repository
  4. The Space auto-deploys on every push

Option 2: Manual Upload

  1. Go to your Space on HuggingFace
  2. Click "Files and versions" → "Add file"
  3. Upload app.py and requirements.txt
  4. The Space builds automatically

Requirements

Your requirements.txt must include:

gradio==4.19.0
plotly==5.18.0
transformers==4.53.0
torch==2.8.0
sentencepiece==0.1.99

Environment Variables

Set these in Space Settings → Variables and secrets:

  • HF_TOKEN (optional): Required for private models
  • MODEL_NAME (optional): Defaults to IberaSoft/customer-sentiment-analyzer

Model Repository Fix

If you encounter "model not recognized" errors, the model repository is missing model_type in config.json.

Fix Using Script

# Get HuggingFace token from: https://huggingface.co/settings/tokens
export HF_TOKEN="your_token_here"

python scripts/fix_model_repo.py \
  --model-id IberaSoft/customer-sentiment-analyzer

This adds the missing model_type and tokenizer files to your model repository.

Troubleshooting

ModuleNotFoundError

Cause: Missing dependency in requirements.txt

Fix: Verify all required packages are listed and restart the Space

Model Loading Fails

Cause: Missing model_type or tokenizer files

Fix: Run the fix_model_repo.py script (see above)

@gr.cache AttributeError

Cause: Using deprecated Gradio decorator

Fix: Remove @gr.cache from your code (not available in Gradio 4.x)

Updating Your Space

Via GitHub

git push origin main

The Space auto-updates within a few minutes.

Via Web Interface

  1. Go to your Space → "Files and versions"
  2. Click on the file to edit
  3. Make changes and commit
  4. Space rebuilds automatically

Testing Locally

Before deploying, test your Gradio app locally:

pip install -r requirements.txt
python app.py

Visit http://localhost:7860 to test the interface.

Verification Checklist

After deployment:

  • Space loads without errors
  • Model loads successfully (check build logs)
  • Can input text and receive predictions
  • Confidence scores display correctly

Common Issues

Space stuck building: Check logs for missing dependencies or syntax errors

Model not found: Verify model is public or HF_TOKEN is set correctly

Gradio version conflicts: Ensure you're using Gradio 4.19.0 or newer