Gahow Wang f26db882e5 Merge t16-grad-accum into main ...

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

# Conflicts:
#	README.md
#	docs/evolution.md

2026-06-18 00:37:11 +08:00

..

xtrain-autodiff

test: eps=2e-3 for flash dQ/dK finite-diff (cuts f32 rounding term)

2026-06-17 23:17:44 +08:00

xtrain-cuda

cuda: fused flash-attention kernel (fwd + flash-style bwd)

2026-06-17 23:10:25 +08:00

xtrain-distributed

Merge t16-grad-accum into main

2026-06-18 00:37:11 +08:00

xtrain-model

test: flash==composed bf16 uses robust mean/p99 metric (repo convention)

2026-06-17 23:19:08 +08:00

xtrain-optim

perf: make xtrain-cuda a regular dep of xtrain-optim (GPU AdamW)

2026-06-15 16:53:52 +08:00

xtrain-tensor

cuda: fused flash-attention kernel (fwd + flash-style bwd)

2026-06-17 23:10:25 +08:00

xtrain-train

Merge t16-grad-accum into main

2026-06-18 00:37:11 +08:00