Skip to content
File size cap (≤500 LoC per .rs)

Lift max_batch_size=1 under EP=2 via head↔worker slot multiplexing (#99) #255

Sign in to view logs

Lift max_batch_size=1 under EP=2 via head↔worker slot multiplexing (#99)

Lift max_batch_size=1 under EP=2 via head↔worker slot multiplexing (#99) #255

Re-run triggered June 1, 2026 13:28
@tbraun96tbraun96
#101
camerono:pr/ep-batched-decode
Status Failure
Total duration 18s
Artifacts –

file-size-cap.yml

on: pull_request
Enforce ≤500 LoC per source file
7s
Enforce ≤500 LoC per source file
Fit to window
Zoom out
Zoom in

Annotations

2 errors and 10 warnings
Enforce ≤500 LoC per source file
Process completed with exit code 1.
Enforce ≤500 LoC per source file: crates/spark-model/src/model/impl_a2.rs#L0
535 LoC exceeds 500-line cap. Split into sub-modules per Atlas idiom (see crates/spark-model/src/layers/qwen3_attention/ for compute-heavy template, crates/spark-model/src/weight_loader/ for variant-dispatch template).
Enforce ≤500 LoC per source file: crates/spark-server/tests/integration.rs#L0
Allow-listed: 509 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-server/src/main_modules/serve.rs#L0
Allow-listed: 537 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-server/src/scheduler/decode_logits_step.rs#L0
Allow-listed: 510 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-comm/src/nccl_backend.rs#L0
Allow-listed: 524 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-model/src/traits/model.rs#L0
Allow-listed: 663 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-model/src/weight_loader/gemma4/loader_a.rs#L0
Allow-listed: 513 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-model/src/model/trait_impl/prefill_c.rs#L0
Allow-listed: 512 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-model/src/layers/qwen3_attention/decode/attention_forward.rs#L0
Allow-listed: 507 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-model/src/layers/qwen3_attention/trait_impl/multi_seq/qkv.rs#L0
Allow-listed: 537 LoC (>500). Reduce when feasible.
Enforce ≤500 LoC per source file: crates/spark-runtime/src/metal_backend.rs#L0
Allow-listed: 733 LoC (>500). Reduce when feasible.