aboutsummaryrefslogtreecommitdiff
Commit message (Expand)AuthorAgeFilesLines
* docs: update README introHEADv0.3.0masterOnur Gündüz2026-07-141-19/+9
* chore: move the server README to top-levelOnur Gündüz2026-07-142-271/+263
* chore: update READMEOnur Gündüz2026-07-141-26/+0
* bench: add tokens-per-text and float embeddings to the server benchOnur Gündüz2026-07-041-6/+24
* server: bound the collect window by the batch token budgetOnur Gündüz2026-07-042-0/+63
* server: per-model input token cap, auto-truncate, and 429 admissionOnur Gündüz2026-07-049-24/+203
* server: make the per-request token cap configurableOnur Gündüz2026-07-025-6/+20
* cuda: default cuDNN attention to Qwen only, BERT opt-inOnur Gündüz2026-07-021-12/+19
* cuda: add cuDNN fused attention path with runtime fallbackOnur Gündüz2026-07-02102-8/+80725
* docs: update READMEOnur Gündüz2026-06-301-1/+2
* docs: update READMEOnur Gündüz2026-06-301-23/+38
* ci: fix the invalid filename in gh release commandv0.2.0Onur Gündüz2026-06-301-1/+1
* ci: add version label to release packagesOnur Gündüz2026-06-301-1/+1
* ci: do not include README in release packagesOnur Gündüz2026-06-301-2/+2
* ci: build and publish CUDA releaseOnur Gündüz2026-06-281-0/+19
* ci: add dry run option to workflowOnur Gündüz2026-06-281-0/+13
* ci: remove yyjson from apt install listOnur Gündüz2026-06-281-2/+0
* docs: fix heading sizeOnur Gündüz2026-06-281-1/+1
* docs: update READMEOnur Gündüz2026-06-281-13/+29
* docs: update server README to clarify GPU precision flagsOnur Gündüz2026-06-271-17/+40
* chore: update README.mdOnur Gündüz2026-06-271-11/+16
* chore: update project name in LICENSEOnur Gündüz2026-06-271-1/+1
* server: add README.mdOnur Gündüz2026-06-271-0/+219
* gitignore: add bench folders to ignore listOnur Gündüz2026-06-271-0/+4
* chore: clean up comment formatting and wordingOnur Gündüz2026-06-2796-746/+792
* cuda: skip Qwen flash softmax rescale when running max is stableOnur Gündüz2026-06-261-8/+33
* cuda: load Qwen flash V with ldmatrix.trans instead of gatherOnur Gündüz2026-06-261-35/+83
* build: vendor yyjson 0.12.0 in deps/yyjsonOnur Gündüz2026-06-2510-17/+19361
* cuda: skip O(L^2) score buffer for qwen flash path (enables 32k)Onur Gündüz2026-06-251-1/+7
* cuda: stage compact GQA K/V in qwen attention, halve the expandOnur Gündüz2026-06-251-14/+47
* chore: fix permissions for test scripts and devtoolsOnur Gündüz2026-06-259-0/+0
* chore: run clang-formatOnur Gündüz2026-06-2520-432/+380
* build: update build info and accelerator labelsOnur Gündüz2026-06-255-34/+30
* server: store the Server-Timing header inline in the jobOnur Gündüz2026-06-252-10/+11
* server: encode base64 embeddings directly into the response bufferOnur Gündüz2026-06-255-67/+97
* server: drop the throwaway buffer in the base64 float pathOnur Gündüz2026-06-251-8/+4
* server: remove the dead auth fieldOnur Gündüz2026-06-252-2/+0
* server: move the request body instead of copying itOnur Gündüz2026-06-251-2/+3
* chore: update READMEOnur Gündüz2026-06-251-13/+20
* all: migrate to yyjsonOnur Gündüz2026-06-2519-323/+354
* cuda: skip per-warp future key tiles in causal qwen attentionOnur Gündüz2026-06-251-65/+95
* cuda: double-buffer qwen flash K/V staging with cp.asyncOnur Gündüz2026-06-241-82/+228
* cuda: prefetch bert flash QK ahead of softmax to overlap the tensor coresOnur Gündüz2026-06-241-50/+99
* cuda: read bert flash K directly, drop the shared-memory transposeOnur Gündüz2026-06-241-40/+65
* cuda: broaden bert flash kv pipeline to all flash batchesOnur Gündüz2026-06-241-8/+10
* tokenizer: parse BPE vocab.json with yyjsonOnur Gündüz2026-06-241-127/+25
* api: parse model config with yyjson instead of a substring scannerOnur Gündüz2026-06-241-77/+94
* build: vendor yyjson 0.12.0 as the core JSON libraryOnur Gündüz2026-06-242-0/+6
* tests: cover CUDA reduced-precision stack and add determinism gateOnur Gündüz2026-06-241-25/+45
* cuda: fix bert reduced-precision ffn gate-up output aliasing its inputOnur Gündüz2026-06-241-129/+156