compute decouple#109
Merged
Merged
Conversation
Resolves async_compute_backend/cpu.rs conflict by taking the HEAD shim — impl lives in larql_compute::async_compute_backend::cpu post-ADR-0022. Brings in #113 (Windows attention_step_async fix).
…rgence
larql-compute-metal: 4 new tests under StateDumpMask::{Full,HOnly,None}
plus the unmasked wrapper. decode/mod.rs 82.57% -> 93.36% — the debt
baseline added yesterday drops; only trait_impl/mod.rs stays at 87
pending a real Q4_K vindex fixture for the hybrid attention body
(synthetic Q4_K bytes null-ptr in decode_attention_layer).
larql-server: temporary diagnostic on the 6 OK||500 completions
tests — strict assert_eq!(status, OK) + body capture via diag_capture.
On macOS aarch64 + Linux x86_64 (Docker/Rosetta) all 12 tests pass
with real responses; CI Ubuntu shows completions.rs at 70.34% line
coverage vs 86.85% local with identical test outcomes. Strict
assertions + body dump in next CI run will surface what the success
path looks like on real x86_64 Ubuntu. Revert after the divergence is
identified.
…time Root cause of the Ubuntu CI vs macOS divergence on completions.rs (70.34% CI / 86.85% local): the `completions_*_returns_200` tests asserted `OK || is_server_error()` on `resp.status()` without draining the response body. axum builds the JSON body lazily via `into_response()`, so on Ubuntu llvm-cov the buffered handler's serialisation tail never ran — ~54 lines of the OK branch in completions.rs stayed uncovered. macOS happened to run them eagerly, masking the issue locally. Fixed by `capture_completion(resp) -> (StatusCode, String)` — drains into a `String` so the response serialisation always runs, then tightening the asserts to strict `StatusCode::OK`. Same 12 tests, now consistent across platforms: completions.rs reports 86.85% everywhere. No synthetic-Q4K generation regression existed; the Linux Docker run (synthetic_vindex_generates_at_least_one_token) confirms generation produces finite logits. Investigation note retained in the policy file. Server policy baseline ratcheted back 70 → 86 (the value the file actually clears now).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
working on decoupling compute, so we can have multuple compute engines