rob thijssen
cc95fe28d9
All checks were successful
build-prerelease / Resolve version stamps (push) Successful in 1m45s
build-prerelease / Build neuron-blackwell (push) Successful in 3m40s
build-prerelease / Build cortex binary (push) Successful in 4m27s
build-prerelease / Package cortex RPM (push) Successful in 1m24s
build-prerelease / Build neuron-ampere (push) Successful in 5m30s
build-prerelease / Build neuron-ada (push) Successful in 5m24s
build-prerelease / Package helexa-neuron-ada RPM (push) Successful in 3m6s
build-prerelease / Package helexa-neuron-ampere RPM (push) Successful in 3m6s
build-prerelease / Package helexa-neuron-blackwell RPM (push) Successful in 3m49s
CI / Format (push) Successful in 35s
build-prerelease / Publish to rpm.lair.cafe (unstable) (push) Successful in 1m7s
CI / Clippy (push) Successful in 2m16s
CI / Test (push) Successful in 4m37s
CI / Build cortex SRPM (push) Has been skipped
CI / Build neuron SRPM (push) Has been skipped
CI / Publish cortex to COPR (push) Has been skipped
CI / Publish neuron to COPR (push) Has been skipped
CI / Bump version in source (push) Has been skipped
run_fused_gating helper consolidates the per-layer gating math: beta = sigmoid(b) g = -exp(a_log) * softplus(a + dt_bias) CUDA path issues a single launch via fused_gdn_gating_cuda; cpu path falls back to the original per-op Rust sequence. Replaces ~10 candle launches per linear-attention layer (sigmoid + 2× to_dtype + exp + neg + broadcast_add + softplus + 2× unsqueeze + broadcast_mul) across both single-GPU and TP forward paths. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>