Vega-Vision-v2

Model mdl-3k8m1q

Back to Models
TU

Vega-Vision-v2

Training

Image classification and object detection model for industrial quality control

ImageCode
Training Progress — Epoch 80/12067%
Created

Mar 11, 2026

Parameters

840M

Accuracy

In Progress

Training Time

Active

Data Sources

Image Dataset Alpha

gs://trinidy-media/images-alpha

890.0K records

Internal NAS Archive

nas://192.168.1.40/ml-archive

340.0K records

Endpoint will be available after training completes

Training Logs

TimestampLevelSourceMessage
06:58:12infonexus-gatewayInference request completed successfully — 284 tokens generated in 42ms
06:57:45infonexus-autoscalerAuto-scaling triggered: replicas 2 → 3 (CPU threshold 78%)
06:55:30warnnexus-monitorResponse latency spike detected: p99 = 156ms (threshold: 150ms)
06:52:18infonexus-lbHealth check passed — all 3 replicas healthy
06:48:02debugnexus-cacheCache hit ratio: 34.2% — model weights warm across all replicas
06:45:11infonexus-gatewayRate limit adjusted: 500 req/s → 750 req/s for endpoint /v1/orion-nlp-v3
06:40:55errornexus-gatewayRequest timeout after 30s — client disconnected (req_id: 8f2a1c)
06:38:22infonexus-storageModel checkpoint synced to cold storage (version 3.1.0)