aboutsummaryrefslogtreecommitdiff
path: root/embed_cuda.cu (follow)
Commit message (Collapse)AuthorAgeFilesLines
* chore: move sources to src/ with proper names, drop qwen_ prefixOnur Gündüz2026-06-161-2052/+0
|
* cuda: memory-only bf16 for BERT (f32 compute on bf16 weights)Onur Gündüz2026-06-161-6/+46
|
* all: support BERT tanh-GeLU (gelu_new / gelu_pytorch_tanh)Onur Gündüz2026-06-161-4/+18
|
* cuda: add the BERT encoder family forward and loaderOnur Gündüz2026-06-161-9/+238
|
* all: add CLS first-token poolingOnur Gündüz2026-06-151-0/+13
|
* all: support GTE-Qwen2 embeddingsOnur Gündüz2026-06-151-17/+55
|
* late: add CUDA late-interaction rerank path (GPU encode, host MaxSim)Onur Gündüz2026-06-151-4/+219
|
* chore: run clang-formatOnur Gündüz2026-06-151-442/+428
|
* all: add support for Qwen3-Embedding + causal GQA attentionOnur Gündüz2026-06-141-10/+91
|
* chore: rename projectOnur Gündüz2026-06-141-55/+55
|
* cuda: restore BF16 attention internals, length-route batched attentionOnur Gündüz2026-06-141-42/+130
|
* cuda: drop BF16 attention internals, keep fused rope + batched F32 attentionOnur Gündüz2026-06-141-125/+42
|
* cuda: pointer-array batched attention for equal-length micro-batchesOnur Gündüz2026-06-141-8/+167
|
* cuda: fuse QK norm+rope launches, BF16 attention output pathOnur Gündüz2026-06-141-45/+99
|
* cuda: follow model file dtype for weight storage by defaultOnur Gündüz2026-06-141-1/+15
|
* chore: rename sources; binaries pplx-embed, pplx-embed-serverOnur Gündüz2026-06-141-2/+2
|
* gitignore: track new binary and library namesOnur Gündüz2026-06-141-0/+1204