8000 Commits · ngxson/llama.cpp · GitHub
[go: up one dir, main page]
Skip to content

Commits

Branch selector

User selector

Datepicker

Commit History

Commits on May 9, 2025

  • CUDA: FA support for Deepseek (Ampere or newer) (#13306)

    JohannesGaesslerJohannesGaessler
    authored
    0cf6725

  • llama : do not crash if there is no CPU backend (#13395)

    slarenslaren
    authored
    27ebfca

  • CUDA: fix crash on large batch size for MoE models (#13384)

    JohannesGaesslerJohannesGaessler
    authored
    5c86c9e

  • imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389)

    bartowski1182bartowski1182
    authored
    efb8b47

  • llama-run: add support for downloading models from ModelScope (#13370)

    yeahdongcnyeahdongcn
    authored
    0527771

  • mtmd : fix batch_view for m-rope (#13397)

    ngxsonngxson
    authored
    2189fd3

  • llama : one-off chat template fix for Mistral-Small-2503 (#13398)

    ngxsonngxson
    authored
    3f96aef

  • rpc : add rpc_msg_set_tensor_hash_req (#13353)

    rgerganovrgerganov
    authored
    b486ba0

  • vulkan: Allow up to 4096 elements for mul_mat_id row_ids (#13326)

    jeffbolznvjeffbolznv
    authored
    02115dc

  • server : (webui) rename has_multimodal --> modalities (#13393)

    ngxsonngxson
    authored
    d9c4acc

    • Commits on May 8, 2025

    Pagination

    0