aboutsummaryrefslogtreecommitdiff
Commit message (Expand)AuthorAgeFilesLines
* cpu: add the BERT encoder family forward and loaderOnur Gündüz2026-06-154-17/+276
* kernels: add LayerNorm and exact-GeLU for the BERT familyOnur Gündüz2026-06-153-0/+105
* tokenizer: add WordPiece backend for BERT-family modelsOnur Gündüz2026-06-154-0/+969
* all: add CLS first-token poolingOnur Gündüz2026-06-159-27/+165
* gitignore: add test_qwen2 and test_qwen3 to ignored filesOnur Gündüz2026-06-151-0/+2
* server: guard empty-chunk memcpy and check self-pipe writesOnur Gündüz2026-06-151-4/+12
* all: support GTE-Qwen2 embeddingsOnur Gündüz2026-06-1511-85/+574
* api: share config parsing across backendsOnur Gündüz2026-06-155-280/+166
* api: accept Hugging Face model weight prefixOnur Gündüz2026-06-156-65/+134
* server: add Qwen3 text_type query/document instruction hintOnur Gündüz2026-06-152-2/+163
* late: add CUDA late-interaction rerank path (GPU encode, host MaxSim)Onur Gündüz2026-06-155-13/+346
* late: restore MLX device select/concat used by interaction checkOnur Gündüz2026-06-152-0/+111
* late: batch CPU rerank candidate encoding in one packed forwardOnur Gündüz2026-06-154-21/+219
* late: batch MLX rerank candidate encoding in one padded passOnur Gündüz2026-06-153-111/+254
* chore: run clang-formatOnur Gündüz2026-06-1535-4024/+4742
* cli: update print format in build infoOnur Gündüz2026-06-151-5/+5
* server: version flags, drop cost block, tidy batch-wait noteOnur Gündüz2026-06-152-37/+35
* cli: add --json output, plain matrix, and version flagsOnur Gündüz2026-06-153-38/+122
* build: backend-selectable targets, server-only cjson, version flagsOnur Gündüz2026-06-151-66/+51
* api: add embed_build.h for version and build identityOnur Gündüz2026-06-151-0/+64
* ci: add github release workflowOnur Gündüz2026-06-152-3/+231
* server: document HTTP API and add openapi.yaml specOnur Gündüz2026-06-152-44/+434
* server: per-family embedding encoding (Qwen3 float, pplx int8)Onur Gündüz2026-06-143-57/+170
* server: add Qwen3-Embedding-8B model slotOnur Gündüz2026-06-142-2/+5
* all: add Qwen3-Embedding-4B + fix terminal tokenOnur Gündüz2026-06-147-25/+29
* chore: add MIT licenseOnur Gündüz2026-06-141-0/+20
* docs: add acknowledgements to READMEOnur Gündüz2026-06-141-0/+15
* all: add support for Qwen3-Embedding + causal GQA attentionOnur Gündüz2026-06-1416-132/+901
* docs: correct path to rank_embeddings.py in usage exampleOnur Gündüz2026-06-141-1/+1
* docs: update README with the model support and usage examplesOnur Gündüz2026-06-141-3/+62
* docs: fix CLI name in commentsOnur Gündüz2026-06-142-5/+2
* cli: scope GPU options to their respective build targetsOnur Gündüz2026-06-143-122/+130
* docs: fix benchstat.py path in commentsOnur Gündüz2026-06-142-3/+3
* server: remove CORS supportOnur Gündüz2026-06-144-16/+5
* cli: remove --backend, use --mlx and --cuda consistentlyOnur Gündüz2026-06-142-66/+11
* build: add gpu target for auto-selecting backendOnur Gündüz2026-06-142-204/+35
* chore: fix static analyzer warnings in tests/benchOnur Gündüz2026-06-143-22/+47
* server: add late-interaction rerank APIOnur Gündüz2026-06-143-48/+721
* server: tune admission deadline by backendOnur Gündüz2026-06-141-2/+5
* server: make microbatch admission deterministicOnur Gündüz2026-06-142-30/+131
* build: add test-safetensors to PHONY targetsOnur Gündüz2026-06-141-2/+2
* build: add test-bf16 and test-safetensors targets to MakefileOnur Gündüz2026-06-141-1/+12
* tools: add tools for model conversion/inspection and ranking embeddingsOnur Gündüz2026-06-143-0/+377
* chore: rename projectOnur Gündüz2026-06-1422-814/+805
* tools: add benchmark comparison scriptOnur Gündüz2026-06-141-0/+127
* chore: comments housekeepingOnur Gündüz2026-06-145-24/+17
* bench: remove the legacy bench scriptOnur Gündüz2026-06-141-500/+0
* cuda: restore BF16 attention internals, length-route batched attentionOnur Gündüz2026-06-141-42/+130
* dist: remove distributed layer-sharding (protocol, CLI flags, slice APIs)Onur Gündüz2026-06-149-1403/+19
* cuda: drop BF16 attention internals, keep fused rope + batched F32 attentionOnur Gündüz2026-06-141-125/+42