Skip to content

Latest commit

 

History

History
29 lines (24 loc) · 898 Bytes

runpod_progress.md

File metadata and controls

29 lines (24 loc) · 898 Bytes

Jake AI Training Progress Log

Current Progress

  • Identified issues with Colab GPU and CUDA setup
  • Decided to move to RunPod for training
  • Selected PyTorch 2.1 template with A5000 GPU ($0.43/hr)

Next Steps

  1. Complete RunPod deployment
  2. Upload training data to Jupyter interface
  3. Set up training script
  4. Configure Hugging Face integration

Cost Summary

  • Training: ~$0.43/hr on RunPod (A5000 GPU)
  • Inference: Free using Hugging Face API
  • Estimated total training cost: $1.72-$2.58 (4-6 hours)

Technical Setup Selected

  • Platform: RunPod
  • GPU: NVIDIA RTX A5000 (24GB VRAM)
  • Container: PyTorch 2.1 (runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04)
  • Base OS: Ubuntu 22.04

Deployment Plan

  1. Use RunPod for model training
  2. Push trained model to Hugging Face
  3. Use Hugging Face's free inference API for deployment
  4. Generate tweets using the deployed model