Training Running at Full Capacity

Training Running at Full Capacity | Generated by AI

Home 2026.06

Training is running fine, not stuck at all.

Status:

Process PID 71496: running at 99% CPU, 83 min uptime
GPU: RTX 4070 at 100% utilization, 5050 MiB VRAM used, 65°C, 208W draw
Script: train.py config/train_github_code_124m.py on nanoGPT

The only output so far are two warnings (deprecated GradScaler API + “not enough SMs for max_autotune_gemm”), but the GPU being pegged at 100% with 5GB VRAM confirms active computation. The inductor/compile step may have been slow initially, but it’s crunching now.

Back Donate