sergio
5351d4a57b
Some checks failed
build_docker / essential (push) Successful in 1s
build_docker / build_paddle_ocr (push) Failing after 4m34s
build_docker / build_easyocr (push) Has been cancelled
build_docker / build_easyocr_gpu (push) Has been cancelled
build_docker / build_doctr (push) Has been cancelled
build_docker / build_doctr_gpu (push) Has been cancelled
build_docker / build_raytune (push) Has been cancelled
build_docker / build_paddle_ocr_gpu (push) Has been cancelled
19 lines
1.1 KiB
Markdown
19 lines
1.1 KiB
Markdown
# Slide 8 — Architecture: Docker Microservices (~1 minute)
|
|
|
|
**What to say:**
|
|
|
|
"Para la infraestructura, diseñamos una arquitectura de microservicios con Docker. Tenemos dos contenedores principales.
|
|
|
|
El contenedor de Ray Tune actúa como orquestador de ensayos, usando Optuna TPE para decidir qué configuración probar en cada iteración.
|
|
|
|
El contenedor de PaddleOCR recibe las configuraciones vía API REST — un POST a /evaluate — y devuelve las métricas CER, WER y tiempo en formato JSON.
|
|
|
|
Todo se despliega con Docker Compose, lo que permite reproducir el experimento con un solo comando. El hardware utilizado fue un portátil con GPU RTX 3060 Laptop de casi 6 gigabytes de VRAM, procesador AMD Ryzen 7, y 16 gigas de RAM, bajo Ubuntu 24.04.
|
|
|
|
Esta separación en microservicios permite escalar independientemente cada componente y facilita la portabilidad a otros entornos."
|
|
|
|
**Tips:**
|
|
- Point to the architecture diagram as you explain the flow
|
|
- Don't get bogged down in Docker details — focus on the WHY (reproducibility, scalability)
|
|
- Mention hardware briefly — it shows this is feasible with consumer-grade equipment
|