Commits

Commit history

1. Commits on Dec 27, 2025

   • llama_fit_params: return enum for fail vs. error (#18374)

     JohannesGaessler

     authored
     a52dc60View commit details

     Copy full SHA for a52dc60

     View code at this pointBrowse repository at this point

   • llama-fit-params: fix Gemma 3 calculation (#18372)

     JohannesGaessler

     authored
     9045c9aView commit details

     Copy full SHA for 9045c9a

     View code at this pointBrowse repository at this point
2. Commits on Dec 24, 2025

   • model: support MiMo-V2-Flash (#18328)

     Show description for 4cbafad
     

     ngxson

     and

     Aaryan-Kapoor

     authored
     4cbafadView commit details

     Copy full SHA for 4cbafad

     View code at this pointBrowse repository at this point

   • model : support for LlamaBidirectionalModel architecture (#18220)

     Show description for 54132f1
     sfallah

     authored
     54132f1View commit details

     Copy full SHA for 54132f1

     View code at this pointBrowse repository at this point
3. Commits on Dec 23, 2025

   • model : fix div-by-zero for Nemotron V2 (#18309)

     Show description for 96e33a8
     

     Alessandro98-git

     and

     CISC

     authored
     96e33a8View commit details

     Copy full SHA for 96e33a8

     View code at this pointBrowse repository at this point
4. Commits on Dec 22, 2025

   • model : Granite Embedding support (#15641)

     Show description for dfc959b
     

     3 peopleauthored
     dfc959bView commit details

     Copy full SHA for dfc959b

     View code at this pointBrowse repository at this point

   • tool/ex/tests: consistently free ctx, then model (#18168)

     JohannesGaessler

     authored
     147a521View commit details

     Copy full SHA for 147a521

     View code at this pointBrowse repository at this point
5. Commits on Dec 19, 2025

   • llama : Changing off_t to size_t for Windows (#18204)

     JTischbein

     authored
     f99ef53View commit details

     Copy full SHA for f99ef53

     View code at this pointBrowse repository at this point
6. Commits on Dec 18, 2025

   • llama: offload output layer to GPU first (#18148)

     JohannesGaessler

     authored
     57c1e05View commit details

     Copy full SHA for 57c1e05

     View code at this pointBrowse repository at this point

   • llama : Async DirectIO model loading on Linux (#18012)

     Show description for 4d4f4ca
     JTischbein

     authored
     4d4f4caView commit details

     Copy full SHA for 4d4f4ca

     View code at this pointBrowse repository at this point
7. Commits on Dec 17, 2025

   • llama-fit-params: fix memory print (#18136)

     JohannesGaessler

     authored
     8dcc366View commit details

     Copy full SHA for 8dcc366

     View code at this pointBrowse repository at this point

   • common : restore grammar-based rejection sampling (#18137)

     Show description for 4301e27
     ggerganov

     authored
     4301e27View commit details

     Copy full SHA for 4301e27

     View code at this pointBrowse repository at this point

   • model: fix LFM2_MOE missing tensors (#18132)

     tdakhran

     authored
     982060fView commit details

     Copy full SHA for 982060f

     View code at this pointBrowse repository at this point
8. Commits on Dec 16, 2025

   • llama-fit-params: force disable mlock (#18103)

     JohannesGaessler

     authored
     d0794e8View commit details

     Copy full SHA for d0794e8

     View code at this pointBrowse repository at this point

   • llama-fit-params: lower ctx size for multi GPU (#18101)

     JohannesGaessler

     authored
     9dcac6cView commit details

     Copy full SHA for 9dcac6c

     View code at this pointBrowse repository at this point

   • llama-fit-params: fix underflow for dense models (#18095)

     JohannesGaessler

     authored
     0e49a7bView commit details

     Copy full SHA for 0e49a7b

     View code at this pointBrowse repository at this point

   • model: fix LFM2 missing tensors (#18105)

     ngxson

     authored
     ef83fb8View commit details

     Copy full SHA for ef83fb8

     View code at this pointBrowse repository at this point

   • llama: fix early stop in params_fit if ctx is set (#18070)

     JohannesGaessler

     authored
     ec98e20View commit details

     Copy full SHA for ec98e20

     View code at this pointBrowse repository at this point

   • arch: refactor LLM_TENSOR_NAMES (#18051)

     Show description for 7f2b2f3
     ngxson

     authored
     7f2b2f3View commit details

     Copy full SHA for 7f2b2f3

     View code at this pointBrowse repository at this point

   • Optimization: Qwen3 next autoregressive pass (#17996)

     Show description for a5251ca
     pwilkin

     authored
     a5251caView commit details

     Copy full SHA for a5251ca

     View code at this pointBrowse repository at this point

   • model: support GLM4V vision encoder (#18042)

     Show description for 3d86c6c
     

     ngxson

     and

     ggerganov

     authored
     3d86c6cView commit details

     Copy full SHA for 3d86c6c

     View code at this pointBrowse repository at this point

   • llama: Include algorithm header needed for C++23 (#18078)

     cpeterso

     authored
     2aa45efView commit details

     Copy full SHA for 2aa45ef

     View code at this pointBrowse repository at this point

   • graph : reuse SSM graphs (#16490)

     Show description for c560316
     ggerganov

     authored
     c560316View commit details

     Copy full SHA for c560316

     View code at this pointBrowse repository at this point

   • llama : add support for NVIDIA Nemotron 3 Nano (#18058)

     Show description for 2995341
     

     danbev

     and

     ggerganov

     authored
     2995341View commit details

     Copy full SHA for 2995341

     View code at this pointBrowse repository at this point
9. Commits on Dec 15, 2025

   • model : add KORMo model (#18032)

     Show description for 9d52f17
     HelloKS

     authored
     9d52f17View commit details

     Copy full SHA for 9d52f17

     View code at this pointBrowse repository at this point

   • kv-cache: Fix state restore fragmented cache (#17982)

     Show description for 4529c66
     

     ssweens

     and

     ggerganov

     authored
     4529c66View commit details

     Copy full SHA for 4529c66

     View code at this pointBrowse repository at this point

   • llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653)

     Show description for b1f3a6e
     JohannesGaessler

     authored
     b1f3a6eView commit details

     Copy full SHA for b1f3a6e

     View code at this pointBrowse repository at this point
10. Commits on Dec 14, 2025

    • graph: add f_attn_temp_offset (#18025)

      ngxson

      authored
      0759b09View commit details

      Copy full SHA for 0759b09

      View code at this pointBrowse repository at this point

    • models : fix YaRN regression + consolidate logic (#18006)

      Show description for 609a2d0
      ggerganov

      authored
      609a2d0View commit details

      Copy full SHA for 609a2d0

      View code at this pointBrowse repository at this point
11. Commits on Dec 13, 2025

    • llama_context: synchronize before reallocating output buffer (#17974)

      jeffbolznv

      authored
      5266379View commit details

      Copy full SHA for 5266379

      View code at this pointBrowse repository at this point
12. Commits on Dec 12, 2025

    • models : fix the attn_factor for mistral3 graphs + improve consistency (#17945)

      Show description for 7bed317
      ggerganov

      authored
      7bed317View commit details

      Copy full SHA for 7bed317

      View code at this pointBrowse repository at this point
13. Commits on Dec 11, 2025

    • batch : fix sequence id ownership (#17915)

      Show description for d9f8f60
      ggerganov

      authored
      d9f8f60View commit details

      Copy full SHA for d9f8f60

      View code at this pointBrowse repository at this point
14. Commits on Dec 10, 2025

    • ggml : remove GGML_KQ_MASK_PAD constant (#17910)

      Show description for 4dff236
      ggerganov

      authored
      4dff236View commit details

      Copy full SHA for 4dff236

      View code at this pointBrowse repository at this point

    • model : Qwen3-Next-80B-A3B has 48 layers (#17898)

      Show description for b677721
      EZForever

      authored
      b677721View commit details

      Copy full SHA for b677721

      View code at this pointBrowse repository at this point
15. Commits on Dec 9, 2025

    • cmake: fix Mach-O current version number (#17877)

      Show description for 63908b6
      Rhys-T

      authored
      63908b6View commit details

      Copy full SHA for 63908b6

      View code at this pointBrowse repository at this point