New pull request
Confirm to delete all selected items?
docs(perf): use idle-card numbers for the GQA attention result
All checks were successful
ci / test_and_build (pull_request) Successful in 41s
#130 by rcsheets was merged 2026-11-04 02:19:49 +00:00
perf(cuda): make decode attention GQA-aware
All checks were successful
ci / test_and_build (pull_request) Successful in 41s
#129 by rcsheets was merged 2026-11-03 08:27:01 +00:00
perf(cuda): tile long decode spans across blocks
All checks were successful
ci / test_and_build (pull_request) Successful in 1m57s
#127 by rcsheets was merged 2026-11-03 04:37:05 +00:00
feat(engine): --power-budget-wait waits out a startup power refusal
All checks were successful
ci / test_and_build (pull_request) Successful in 2m51s
#121 by rcsheets was merged 2026-10-23 01:59:00 +00:00
docs: energy metering at inference time
All checks were successful
ci / test_and_build (pull_request) Successful in 50s
#120 by rcsheets was merged 2026-10-09 23:59:23 +00:00
docs(readme): thinking-budget guidance measured on four models
All checks were successful
ci / test_and_build (pull_request) Successful in 48s
#119 by rcsheets was merged 2026-10-10 00:27:07 +00:00
feat(tokenizer): OLMo 3 Think chat template
All checks were successful
ci / test_and_build (pull_request) Successful in 48s
#118 by rcsheets was merged 2026-10-04 06:28:35 +00:00
feat(model): OLMo 3 (Olmo3ForCausalLM) in the dense-transformer package
All checks were successful
ci / test_and_build (pull_request) Successful in 56s
#117 by rcsheets was merged 2026-10-04 06:25:29 +00:00
feat(backend): sliding-window PagedAttention
All checks were successful
ci / test_and_build (pull_request) Successful in 51s
#116 by rcsheets was merged 2026-10-04 06:23:27 +00:00
feat(config): v4 rope_scaling, per-layer-type rope, and layer_types
All checks were successful
ci / test_and_build (pull_request) Successful in 46s
#115 by rcsheets was merged 2026-10-04 06:15:28 +00:00
feat: stop on generation_config's eos_token_id as well as the tokenizer's EOS
All checks were successful
ci / test_and_build (pull_request) Successful in 1m1s
#114 by rcsheets was merged 2026-10-04 06:11:19 +00:00
refactor: reasoning markers that are text, not one token
All checks were successful
ci / test_and_build (pull_request) Successful in 51s
#113 by rcsheets was merged 2026-10-04 05:11:30 +00:00
perf(engine): sample without sorting the vocabulary, rows side by side
All checks were successful
ci / test_and_build (pull_request) Successful in 55s
#112 by rcsheets was merged 2026-10-03 20:18:18 +00:00
docs: knobs for the operator's goals, and measuring what they do
All checks were successful
ci / test_and_build (pull_request) Successful in 30s
#111 by rcsheets was merged 2026-10-03 20:31:09 +00:00
feat: soft and hard thinking-token budgets for reasoning turns
All checks were successful
ci / test_and_build (pull_request) Successful in 30s
#110 by rcsheets was merged 2026-10-03 18:12:12 +00:00
fix(engine): apply a response_format grammar to the answer, not the reasoning
All checks were successful
ci / test_and_build (pull_request) Successful in 34s
#109 by rcsheets was merged 2026-10-02 19:57:52 +00:00
feat(engine): name the served model on each perfstats line
All checks were successful
ci / test_and_build (pull_request) Successful in 30s
#108 by rcsheets was merged 2026-10-02 19:58:35 +00:00
feat(engine): response debug preview, unlimited previews in files, and closing lines for unfinished requests
All checks were successful
ci / test_and_build (pull_request) Successful in 34s
#107 by rcsheets was merged 2026-10-02 08:17:05 +00:00
test(cuda): add cold-weight and skewed cases to the grouped GEMV benchmark
All checks were successful
ci / test_and_build (pull_request) Successful in 27s
#106 by rcsheets was merged 2026-10-02 08:12:13 +00:00
perf(engine): read logits into a reused buffer and pick greedy tokens in parallel
All checks were successful
ci / test_and_build (pull_request) Successful in 36s
#105 by rcsheets was merged 2026-10-02 07:28:39 +00:00