aboutsummaryrefslogtreecommitdiff
path: root/src/forward.c (unfollow)
Commit message (Collapse)AuthorFilesLines
2026-06-21all: split Makefile and reorg library/toolsOnur Gündüz1-710/+0
2026-06-16all: add SentencePiece XLM-R supportOnur Gündüz1-6/+9
2026-06-16cpu: split embed.c into model.c and forward.cOnur Gündüz1-639/+8
2026-06-16cpu: extract late-interaction model into late.cOnur Gündüz1-387/+30
2026-06-16cpu: extract alloc.{c,h} and vec.c from embed.cOnur Gündüz1-122/+4
2026-06-16api: dedup scalar bf16-to-f32 into dtype.hOnur Gündüz1-10/+9
2026-06-16chore: move sources to src/ with proper names, drop qwen_ prefixOnur Gündüz1-35/+34
2026-06-16all: support BERT tanh-GeLU (gelu_new / gelu_pytorch_tanh)Onur Gündüz1-1/+4
2026-06-15cpu: add the BERT encoder family forward and loaderOnur Gündüz1-16/+229
2026-06-15all: add CLS first-token poolingOnur Gündüz1-4/+10
2026-06-15all: support GTE-Qwen2 embeddingsOnur Gündüz1-4/+29
2026-06-15api: share config parsing across backendsOnur Gündüz1-146/+2
2026-06-15api: accept Hugging Face model weight prefixOnur Gündüz1-28/+38
2026-06-15late: add CUDA late-interaction rerank path (GPU encode, host MaxSim)Onur Gündüz1-0/+10
2026-06-15late: batch CPU rerank candidate encoding in one packed forwardOnur Gündüz1-0/+92
2026-06-15chore: run clang-formatOnur Gündüz1-516/+532
2026-06-14all: add Qwen3-Embedding-4B + fix terminal tokenOnur Gündüz1-6/+1
2026-06-14all: add support for Qwen3-Embedding + causal GQA attentionOnur Gündüz1-26/+85
2026-06-14chore: rename projectOnur Gündüz1-146/+146
2026-06-14dist: remove distributed layer-sharding (protocol, CLI flags, slice APIs)Onur Gündüz1-42/+0
2026-06-14late: batch MaxSim as grouped GEMM plus row-max reductionOnur Gündüz1-2/+60
2026-06-14cpu: move BF16 widening scratch into the workspaceOnur Gündüz1-18/+49
2026-06-14chore: rename sources; binaries pplx-embed, pplx-embed-serverOnur Gündüz1-2/+2
2026-06-14gitignore: track new binary and library namesOnur Gündüz1-0/+0
2026-06-14kernels: drop asr_ prefix namingOnur Gündüz1-2/+2
2026-06-14cuda: add initial CUDA embedding backendOnur Gündüz1-57/+1
2026-06-14chore: address static analyzer warningsOnur Gündüz1-2/+5
2026-06-14late: add packed maxsim scoringOnur Gündüz1-3/+22
2026-06-14late: add CPU token vectors and maxsimOnur Gündüz1-5/+188
2026-06-14all: add validation for late-interaction model compatOnur Gündüz1-1/+25
2026-06-14cpu: reuse tiled attention scratch buffersOnur Gündüz1-3/+23
2026-06-14server: batch contextual document inferenceOnur Gündüz1-12/+69
2026-06-14dist: load distributed layer shards selectivelyOnur Gündüz1-14/+46
2026-06-14dist: add distributed layer-slice foundationOnur Gündüz1-21/+113
2026-06-14api: add opt-in l2 normalization helperOnur Gündüz1-3/+31
2026-06-14server: align embedding quantization with perplexityOnur Gündüz1-23/+5
2026-06-14server: implement HTTP API with libae integrationOnur Gündüz1-0/+62
2026-06-14cpu: keep batch offsets in workspaceOnur Gündüz1-10/+38
2026-06-14cpu: tighten workspace growthOnur Gündüz1-4/+14
2026-06-14cpu: reduce initial workspace capacityOnur Gündüz1-1/+2
2026-06-14server: report workspace memoryOnur Gündüz1-0/+58
2026-06-14kernels,cpu: fuse tiny BF16 gate up projectionsOnur Gündüz1-3/+26
2026-06-14kernels,cpu: fuse small BF16 QKV projectionsOnur Gündüz1-3/+24
2026-06-14cpu: reuse MLP scratch bufferOnur Gündüz1-6/+3
2026-06-14cpu: support BF16 weightsOnur Gündüz1-46/+142
2026-06-14cpu: fuse CPU embedding hot pathsOnur Gündüz1-11/+23
2026-06-14cpu: validate model tensor shapesOnur Gündüz1-21/+90
2026-06-14cpu: harden workspace allocationOnur Gündüz1-41/+125
2026-06-14api: simplify embedding APIOnur Gündüz1-80/+46
2026-06-14api: split model and workspace ownershipOnur Gündüz1-98/+203