feat(training): GPU-resident fused step for non-TS single-net models#1843

Merged

ooples merged 17 commits into

masterfrom

feat/gpu-resident-nonts

Jul 11, 2026

Commits on Jul 10, 2026

feat(training): gPU-resident fused step for non-TS single-net models

franklinic
and
claude
committed
feat(training): gPU-resident fused step for Finance foundation forecasters

franklinic
and
claude
committed
feat(training): gPU-resident fused generator step for GAN-family synthetic generators + TVAE

franklinic
and
claude
committed
feat(training): gPU-resident fused step for TabSyn VAE + MisGAN mask/data gens

franklinic
and
claude
committed
feat(training): gPU-resident fused step for TimeGAN P1/P2 + MedSynth non-DP disc

franklinic
and
claude
committed
feat(training): extra-tensors parameter on fused-step API

franklinic
and
claude
committed
feat(training): wire GOGGLE + CLAP through fused-step extra-tensors API

franklinic
and
claude
committed
feat(diffusion): batched per-element timesteps for DDPM-canonical training

franklinic
and
claude
committed
feat(training): wGAN-GP correctness fix + fused disc + DP-SGD wire-ups (4F/4G/4H)

franklinic
and
claude
committed
fix(review): resolve 12 CodeRabbit findings on PR #1843

franklinic
and
claude
committed
feat(training): vectorized fused DP-SGD/WGAN-GP/multi-slot primitives + consumer wire-ups

franklinic
and
claude
committed
Merge branch 'master' into feat/gpu-resident-nonts
ooples
authored
fix(review): tFC + TOTEM preprocessing rewritten with traceable engine ops

franklinic
and
claude
committed
feat(training): centralize WGAN-GP + diffusion consumers through fused primitives (#1847)

authored
feat(training): wire additional WGAN-GP + diffusion consumers through fused primitives (#1848)

authored

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(training): GPU-resident fused step for non-TS single-net models#1843

feat(training): GPU-resident fused step for non-TS single-net models#1843
ooples merged 17 commits into
masterfrom
feat/gpu-resident-nonts

Commits on Jul 10, 2026

feat(training): gPU-resident fused step for non-TS single-net models

feat(training): gPU-resident fused step for Finance foundation forecasters

feat(training): gPU-resident fused generator step for GAN-family synthetic generators + TVAE

feat(training): gPU-resident fused step for TabSyn VAE + MisGAN mask/data gens

feat(training): gPU-resident fused step for TimeGAN P1/P2 + MedSynth non-DP disc

feat(training): extra-tensors parameter on fused-step API

feat(training): wire GOGGLE + CLAP through fused-step extra-tensors API

feat(diffusion): batched per-element timesteps for DDPM-canonical training

feat(training): wGAN-GP correctness fix + fused disc + DP-SGD wire-ups (4F/4G/4H)

fix(review): resolve 12 CodeRabbit findings on PR #1843

feat(training): vectorized fused DP-SGD/WGAN-GP/multi-slot primitives + consumer wire-ups

Merge branch 'master' into feat/gpu-resident-nonts

fix(review): tFC + TOTEM preprocessing rewritten with traceable engine ops

feat(training): centralize WGAN-GP + diffusion consumers through fused primitives (#1847)

feat(training): wire additional WGAN-GP + diffusion consumers through fused primitives (#1848)

Commits on Jul 11, 2026

chore(deps): bump AiDotNet.Tensors 0.111.2 -> 0.113.0 (activates #763 for fused paths)

Merge remote-tracking branch 'origin/master' into feat/gpu-resident-nonts