| Age | Commit message (Expand) | Author |
|---|
| 2024-01-05 | ggml : do not sched_yield when calling BLAS (#4761) | Georgi Gerganov |
| 2024-01-03 | ggml : extend ggml_get_rows, ggml_repeat, ggml_concat (ggml/639) | Guillaume Wenzek |
| 2023-12-30 | ggml : add ggml_cpu_has_avx_vnni() (#4589) | automaticcat |
| 2023-12-29 | ggml : fix some mul mat cases + add tests for src1 F16 (ggml/669) | bssrdf |
| 2023-12-26 | cuda : fix vmm pool with multi GPU (#4620) | slaren |
| 2023-12-26 | Update comment for AdamW implementation reference. (#4604) | WillCorticesAI |
| 2023-12-24 | cuda : improve cuda pool efficiency using virtual memory (#4606) | slaren |
| 2023-12-22 | llama : fix platforms without mmap (#4578) | slaren |
| 2023-12-22 | ggml : add comment about backward GGML_OP_DIAG_MASK_INF (#4203) | Herman Semenov |
| 2023-12-21 | ggml : change ggml_scale to take a float instead of tensor (#4573) | Georgi Gerganov |
| 2023-12-21 | llama : initial ggml-backend integration (#4520) | slaren |
| 2023-12-18 | llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) | Ebey Abraham |
| 2023-12-15 | ggml : group mul_mat_id rows by matrix (cpu only) (#4480) | slaren |
| 2023-12-14 | ggml : use ggml_row_size where possible (#4472) | slaren |
| 2023-12-14 | ggml : remove n_dims from ggml_tensor (#4469) | slaren |
| 2023-12-14 | ggml : add ggml_row_size() (fixes llama out of space) (#4461) | LostRuins |
| 2023-12-14 | ggml : fix OpenCL broadcast requirement for ggml_mul (close #4453) | Georgi Gerganov |
| 2023-12-13 | sync : ggml (SD ops, tests, kernels) (#4444) | Georgi Gerganov |
| 2023-12-13 | llama : add Mixtral support (#4406) | slaren |
| 2023-12-12 | english : use `typos` to fix comments and logs (#4354) | Richard Kiss |
| 2023-12-07 | sync : ggml (new ops, tests, backend, etc.) (#4359) | Georgi Gerganov |
| 2023-12-03 | ggml : reuse ggml_get_n_tasks() in ggml_graph_plan() (#4308) | Georgi Gerganov |
| 2023-12-03 | ggml : fix soft max out-of-bounds access (#4307) | Georgi Gerganov |
| 2023-12-01 | ggml : add ggml_soft_max_ext (#4256) | Georgi Gerganov |
| 2023-11-28 | ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offloa... | Georgi Gerganov |
| 2023-11-26 | ggml : fix -Warray-bounds warning with gcc (#4231) | Jared Van Bortel |
| 2023-11-17 | llama : add functions to get the model's metadata (#4013) | slaren |
| 2023-11-17 | finetune : speed-up ggml_compute_forward_out_prod_f32 via BLAS (#4079) | gwjr |
| 2023-11-16 | gguf : fix potential infinite loops while parsing (#4100) | texmex76 |
| 2023-11-13 | ggml : sync (im2col, GPU conv, 32-bit arm compat) (#4060) | Georgi Gerganov |
| 2023-11-13 | sync : ggml (backend v2) (#3912) | Georgi Gerganov |
| 2023-11-07 | ggml : fix backward rope after YaRN (#3974) | xaedes |
| 2023-11-02 | gguf : print error for GGUFv1 files (#3908) | Georgi Gerganov |
| 2023-11-02 | gguf : remove special-case code for GGUFv1 (#3901) | Georgi Gerganov |
| 2023-11-01 | llama : implement YaRN RoPE scaling (#2268) | cebtenzzre |
| 2023-11-01 | finetune : add -ngl parameter (#3762) | Andrew Godfrey |
| 2023-10-30 | ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861) | Georgi Gerganov |
| 2023-10-29 | ggml : quantization refactoring (#3833) | Georgi Gerganov |
| 2023-10-24 | sync : ggml (conv ops + cuda MSVC fixes) (#3765) | Georgi Gerganov |
| 2023-10-24 | cuda : add batched cuBLAS GEMM for faster attention (#3749) | Georgi Gerganov |
| 2023-10-20 | gguf : support big endian platform (#3552) | Qin Yue Chen |
| 2023-10-20 | ggml : fix rope + llama minor optimizations (#3560) | Herman Semenov |
| 2023-10-13 | ggml : add context enumeration functions (#3605) | slaren |
| 2023-10-12 | examples: support LLaVA v1.5 (multimodal model) (#3436) | M. Yusuf Sarıgöz |
| 2023-10-10 | llm : add MPT support (#3417) | Jan Ploski |
| 2023-10-09 | refact : fix convert script + zero out KV cache to avoid nans (#3523) | Georgi Gerganov |
| 2023-10-08 | sync : ggml (ggml-backend) (#3548) | Georgi Gerganov |
| 2023-10-04 | ggml : fix build after #3329 | Georgi Gerganov |
| 2023-10-04 | llm : add Refact model (#3329) | ds5t5 |
| 2023-10-04 | sync : ggml (conv 1d + 2d updates, UB fixes) (#3468) | Georgi Gerganov |
