Gahow Wang

gahow pushed to feat/two-stop at gahow/aituner

2026-06-15 09:12:35 +00:00

0b6beafeb8 Phase 5: widen search.high to 1.0 to force multi-iteration Stop-B convergence

gahow pushed to feat/two-stop at gahow/aituner

2026-06-15 09:05:41 +00:00

d4aff81691 Add Stop-B end-to-end config (agentic loop, Stop-A enabled)

gahow pushed to main at gahow/xtrain

2026-06-15 09:00:38 +00:00

5e8add2a41 docs: Phase T7 — performance

gahow pushed to feat/two-stop at gahow/aituner

2026-06-15 08:57:55 +00:00

f31e9ccfd5 Record Stop-A boundary-guard A/B: correct verdict, ~38% replay saved

gahow pushed to main at gahow/xtrain

2026-06-15 08:56:18 +00:00

a842e432b5 perf: streams / drop per-op sync

gahow pushed to main at gahow/xtrain

2026-06-15 08:53:55 +00:00

8070c1949a perf: make xtrain-cuda a regular dep of xtrain-optim (GPU AdamW)

gahow pushed to main at gahow/xtrain

2026-06-15 08:53:13 +00:00

b0e397ca81 perf: GPU AdamW + grad-norm

gahow pushed to main at gahow/xtrain

2026-06-15 08:48:46 +00:00

0e5c7d22e2 perf: cuBLAS matmul fwd/bwd

gahow pushed to main at gahow/xtrain

2026-06-15 08:41:13 +00:00

5df1d4d57b test: resolve real_training corpus default via CARGO_MANIFEST_DIR

gahow pushed to main at gahow/xtrain

2026-06-15 08:34:22 +00:00

2f8118fda9 test: tighten AdamW parity (f32 reference, 10 steps, allclose tol)

gahow pushed to main at gahow/xtrain

2026-06-15 08:30:58 +00:00

29b4d30b6c docs: Phase T6 — training loop

22b7434b23 test: AdamW PyTorch parity + checkpoint round-trip + real training

77a82bfeee train: loop + checkpoint save/load + sampler + train binary

7d84a64f5c data: gpt2 bpe via xserv-tokenizer + TinyStories corpus + lr schedule + grad clip

f22429f5b8 optim: hand-written AdamW (decoupled weight decay + bias correction)

Compare 5 commits »

gahow pushed to feat/two-stop at gahow/aituner

2026-06-15 08:27:43 +00:00

03e556f0ab Add Stop-A ON config (adaptive_stop enabled + boundary guard) for A/B

dfc823f972 Add Stop-A SLO-boundary guard

Compare 2 commits »

gahow pushed to main at gahow/xtrain

2026-06-15 08:09:32 +00:00

8565565647 docs: Phase T5 — tiny transformer

603c85e1e0 model: silence torch parity warning (read loss before backward)

Compare 2 commits »

gahow pushed to main at gahow/xtrain

2026-06-15 08:07:32 +00:00

3366f30c4d model: PyTorch parity harness (weight dump + equivalent torch model)

gahow pushed to main at gahow/xtrain

2026-06-15 08:05:27 +00:00

e3912c2380 model: tiny RoPE+RMSNorm+SwiGLU transformer + overfit test

0acfa5df11 ops: grad-check the T5 structural ops

7fb1a29057 ops: embedding/reshape/transpose/split-merge-heads fwd+bwd

Compare 3 commits »

gahow pushed to feat/two-stop at gahow/aituner

2026-06-15 08:03:18 +00:00

9f52812753 Document Stop-A validation: calibration + GPU fidelity check

gahow pushed to main at gahow/xtrain

2026-06-15 07:54:23 +00:00

777f3c7949 docs: Phase T4 — autograd engine

e7ce504b1f ops: differentiable autograd nodes + per-op grad-check tests

Compare 2 commits »

gahow pushed to main at gahow/xtrain

2026-06-15 07:52:36 +00:00

59e33f60e3 docs: Phase T4 — autograd engine

f3a1188f0e ops: differentiable autograd nodes + per-op grad-check tests

Compare 2 commits »

gahow pushed to main at gahow/xtrain

2026-06-15 07:46:52 +00:00

2e68942032 docs: Phase T4 — autograd engine

7de172a2cf ops: differentiable autograd nodes + per-op grad-check tests

224f750ee4 autograd: tape engine + grad accumulation

5aef3742d6 ops: transformer op fwd/bwd CUDA kernels + Tensor wrappers

Compare 4 commits »

gahow pushed to main at gahow/xtrain

2026-06-15 07:29:09 +00:00

88fbe0a85d gemm: realistic f32 tolerances in GEMM acceptance tests