TAPIS · FINE-TUNING

FLEXSERV · PRIVATE MODEL & INFERENCE LAYER

▣ USER-PRIVATE

JUPYTERHUB · USER NOTEBOOK ENVIRONMENT

Fine-tuning App

• Loads the base model
• Trains on user datasets
• Saves tuned weights

Training Dataset

▧images

≡annotations

Private Model Pool

FlexServ storage on users’ $SCRTCH

yolo26n

yolo26n-fine-tuned

API

YOLO Inference API

Runs yolo26n-fine-tuned for every test image

/v1/responses

Responses API

Vibe coding endpoint

for image in test:
  infer(image)

Evaluation Code

Generated by FlexServ

Test Set

▧images

≡labels

prediction = flexserv.infer(image)
metrics.update(prediction, label)

Evaluation Notebook

Compares FlexServ predictions with actual labels

accuracy · precision · recall · mAP

✓

Model Metrics

performance report for the tuned model