| Commit message (Collapse) | Author | Files | Lines | ||
|---|---|---|---|---|---|
| 2026-06-16 | chore: move sources to src/ with proper names, drop qwen_ prefix | Onur Gündüz | 1 | -2752/+0 | |
| 2026-06-16 | all: support BERT tanh-GeLU (gelu_new / gelu_pytorch_tanh) | Onur Gündüz | 1 | -1/+42 | |
| 2026-06-16 | mlx: upcast BERT norm/bias/embedding fields for BF16 models | Onur Gündüz | 1 | -0/+13 | |
| 2026-06-15 | mlx: add the BERT encoder family forward and loader | Onur Gündüz | 1 | -7/+269 | |
| 2026-06-15 | all: add CLS first-token pooling | Onur Gündüz | 1 | -12/+15 | |
| 2026-06-15 | all: support GTE-Qwen2 embeddings | Onur Gündüz | 1 | -8/+40 | |
| 2026-06-15 | api: share config parsing across backends | Onur Gündüz | 1 | -125/+2 | |
| 2026-06-15 | api: accept Hugging Face model weight prefix | Onur Gündüz | 1 | -17/+26 | |
| 2026-06-15 | late: restore MLX device select/concat used by interaction check | Onur Gündüz | 1 | -0/+101 | |
| 2026-06-15 | late: batch MLX rerank candidate encoding in one padded pass | Onur Gündüz | 1 | -74/+216 | |
| 2026-06-15 | chore: run clang-format | Onur Gündüz | 1 | -543/+610 | |
| 2026-06-14 | all: add Qwen3-Embedding-4B + fix terminal token | Onur Gündüz | 1 | -7/+1 | |
| 2026-06-14 | all: add support for Qwen3-Embedding + causal GQA attention | Onur Gündüz | 1 | -25/+126 | |
| 2026-06-14 | chore: rename project | Onur Gündüz | 1 | -103/+103 | |
| 2026-06-14 | dist: remove distributed layer-sharding (protocol, CLI flags, slice APIs) | Onur Gündüz | 1 | -186/+0 | |
| 2026-06-14 | chore: rename sources; binaries pplx-embed, pplx-embed-server | Onur Gündüz | 1 | -3/+3 | |
| 2026-06-14 | gitignore: track new binary and library names | Onur Gündüz | 1 | -0/+0 | |
| 2026-06-14 | kernels: drop asr_ prefix naming | Onur Gündüz | 1 | -1/+1 | |
| 2026-06-14 | mlx: mixed-precision BF16 inference (F32 residual, BF16 weights) | Onur Gündüz | 1 | -7/+71 | |
| 2026-06-14 | mlx: split ragged contextual batches | Onur Gündüz | 1 | -3/+137 | |
| 2026-06-14 | mlx: tune ragged grouping for 4B | Onur Gündüz | 1 | -5/+34 | |
| 2026-06-14 | mlx: prepare contiguous linear weights | Onur Gündüz | 1 | -4/+21 | |
| 2026-06-14 | mlx: optimize ragged embedding batches | Onur Gündüz | 1 | -2/+114 | |
| 2026-06-14 | chore: address static analyzer warnings | Onur Gündüz | 1 | -5/+13 | |
| 2026-06-14 | mlx: use prefix pooling for f32 contextual spans | Onur Gündüz | 1 | -32/+138 | |
| 2026-06-14 | mlx: cache transposed linear weights | Onur Gündüz | 1 | -4/+46 | |
| 2026-06-14 | mlx: keep late maxsim device resident | Onur Gündüz | 1 | -25/+378 | |
| 2026-06-14 | late: add mlx token vector support | Onur Gündüz | 1 | -8/+211 | |
| 2026-06-14 | all: add validation for late-interaction model compat | Onur Gündüz | 1 | -0/+23 | |
| 2026-06-14 | mlx: add opt-in 8-bit weight quantization | Onur Gündüz | 1 | -21/+206 | |
| 2026-06-14 | mlx: harden model and buffer validation | Onur Gündüz | 1 | -46/+156 | |
| 2026-06-14 | mlx: simplify equal-length embedding pooling | Onur Gündüz | 1 | -24/+35 | |
| 2026-06-14 | mlx: pool contextual bf16 spans before widening | Onur Gündüz | 1 | -11/+9 | |
| 2026-06-14 | server: batch contextual document inference | Onur Gündüz | 1 | -35/+119 | |
| 2026-06-14 | mlx: pool contextual spans on device | Onur Gündüz | 1 | -17/+95 | |
| 2026-06-14 | dist: load distributed layer shards selectively | Onur Gündüz | 1 | -15/+56 | |
| 2026-06-14 | dist: add distributed layer-slice foundation | Onur Gündüz | 1 | -207/+255 | |
| 2026-06-14 | server: align embedding quantization with perplexity | Onur Gündüz | 1 | -10/+4 | |
| 2026-06-14 | server: implement HTTP API with libae integration | Onur Gündüz | 1 | -0/+156 | |
| 2026-06-14 | kernels: validate real BF16 artifacts | Onur Gündüz | 1 | -6/+25 | |
| 2026-06-14 | api: simplify embedding API | Onur Gündüz | 1 | -1/+1 | |
| 2026-06-14 | all: support batched embedding attention | Onur Gündüz | 1 | -59/+137 | |
| 2026-03-28 | Refactor and enhance pplx_embed and pplx_embed_mlx for improved model support | Onur Gündüz | 1 | -270/+208 | |
| - Updated `pplx_embed.h` to support new embedding models: `pplx-embed-v1-4b` and `pplx-embed-context-v1-4b`. - Simplified model configuration structure and constants for better maintainability. - Enhanced comments for clarity on model architecture and configuration. - Modified `pplx_embed_mlx.c` to streamline loading of model weights and configuration from `config.json`. - Improved memory management and error handling in the MLX backend. - Updated embedding computation to support dynamic hidden sizes based on model configuration. - Added `pplx_mlx_config` function to retrieve model configuration details. - Updated README.md - Added .gitignore | |||||
| 2026-03-28 | Init | Onur Gündüz | 1 | -0/+465 | |
