Commits

Commit history

1. Commits on May 17, 2026

   • Bump version to 0.3.39

     Show description for a778c57
     JamePeng

     committed
     a778c57View commit details

     Copy full SHA for a778c57

     View code at this pointBrowse repository at this point

   • fix(context): prevent operations on uninitialized or closed contexts

     Show description for 50627bf
     JamePeng

     committed
     50627bfView commit details

     Copy full SHA for 50627bf

     View code at this pointBrowse repository at this point

   • Sync llama.cpp API 20260517

     Show description for 1278812
     JamePeng

     committed
     1278812View commit details

     Copy full SHA for 1278812

     View code at this pointBrowse repository at this point

   • fix(_internals): Remove unnecessary free operations; models should not be released within the context.

     Show description for 39785d0
     JamePeng

     committed
     39785d0View commit details

     Copy full SHA for 39785d0

     View code at this pointBrowse repository at this point
2. Commits on May 16, 2026

   • Bump version to 0.3.39-preview

     Show description for a8f928c
     JamePeng

     committed
     a8f928cView commit details

     Copy full SHA for a8f928c

     View code at this pointBrowse repository at this point

   • feat(core): support loading GGML_BACKEND_DL dynamic backend libraries from wheel lib

     Show description for 038a953
     JamePeng

     committed
     038a953View commit details

     Copy full SHA for 038a953

     View code at this pointBrowse repository at this point

   • fix(_ggml): correct `ggml_backend_unload` function name

     JamePeng

     committed
     6af3cd7View commit details

     Copy full SHA for 6af3cd7

     View code at this pointBrowse repository at this point

   • Implement `Qwen3ASRChatHandler` for `Qwen3-ASR` models.

     Show description for 4d3e320
     JamePeng

     committed
     4d3e320View commit details

     Copy full SHA for 4d3e320

     View code at this pointBrowse repository at this point
3. Commits on May 14, 2026

   • feat(core): integrate fine-grained logging API into Llama class

     Show description for f64320d
     JamePeng

     committed
     f64320dView commit details

     Copy full SHA for f64320d

     View code at this pointBrowse repository at this point

   • feat(logger): refactor and enhance ggml logging configuration system

     Show description for 1fb6a66
     JamePeng

     committed
     1fb6a66View commit details

     Copy full SHA for 1fb6a66

     View code at this pointBrowse repository at this point
4. Commits on May 13, 2026

   • fix(MTMDChatHandler): correct audio_url content type check and improve variable handling

     Show description for 9954393
     JamePeng

     committed
     9954393View commit details

     Copy full SHA for 9954393

     View code at this pointBrowse repository at this point

   • Implement `MiniCPMV46ChatHandler` for `MiniCPM-V-4.6`

     Show description for e67169d
     JamePeng

     committed
     e67169dView commit details

     Copy full SHA for e67169d

     View code at this pointBrowse repository at this point

   • Sync llama.cpp API 20260513

     Show description for 89e90a7
     JamePeng

     committed
     89e90a7View commit details

     Copy full SHA for 89e90a7

     View code at this pointBrowse repository at this point
5. Commits on May 5, 2026

   • feat(HybridCheckpointCache): add `on-device` hybrid checkpoint support

     Show description for 8eafd9e
     JamePeng

     committed
     8eafd9eView commit details

     Copy full SHA for 8eafd9e

     View code at this pointBrowse repository at this point

   • Sync llama : add option to save memory in device buffers

     Show description for 3796562
     JamePeng

     committed
     3796562View commit details

     Copy full SHA for 3796562

     View code at this pointBrowse repository at this point
6. Commits on May 4, 2026

   • Bump version to 0.3.38

     Show description for ef27f33
     JamePeng

     committed
     ef27f33View commit details

     Copy full SHA for ef27f33

     View code at this pointBrowse repository at this point

   • patch(logger): filter out verbose noisy CUDA Graph debug logs

     Show description for 22a7fdb
     JamePeng

     committed
     22a7fdbView commit details

     Copy full SHA for 22a7fdb

     View code at this pointBrowse repository at this point

   • perf: Optimize detokenize buffer sizing for CJK-heavy outputs

     Show description for a63f60e
     JamePeng

     committed
     a63f60eView commit details

     Copy full SHA for a63f60e

     View code at this pointBrowse repository at this point
7. Commits on May 3, 2026

   • feat(grammar): sync JSON schema to GBNF converter with upstream

     Show description for 5336947
     JamePeng

     committed
     5336947View commit details

     Copy full SHA for 5336947

     View code at this pointBrowse repository at this point
8. Commits on May 2, 2026

   • Bump version to 0.3.37

     Show description for fe38cbf
     JamePeng

     committed
     fe38cbfView commit details

     Copy full SHA for fe38cbf

     View code at this pointBrowse repository at this point
9. Commits on Apr 30, 2026

   • feat(llama): add fine-grained MoE CPU offloading controls

     Show description for b7064d7
     JamePeng

     committed
     b7064d7View commit details

     Copy full SHA for b7064d7

     View code at this pointBrowse repository at this point

   • feat(_ggml): implement ggml-backend API bindings and fix type hints

     Show description for 66dd88b
     JamePeng

     committed
     66dd88bView commit details

     Copy full SHA for 66dd88b

     View code at this pointBrowse repository at this point
10. Commits on Apr 27, 2026

    • feat(handler): Support `add_generation_prompt` parameter pass to MTMDChatHandler

      Show description for 969b45f
      JamePeng

      committed
      969b45fView commit details

      Copy full SHA for 969b45f

      View code at this pointBrowse repository at this point
11. Commits on Apr 26, 2026

    • feat(core): implement thread-safe generation abort mechanism

      Show description for 2cccd2e
      JamePeng

      committed
      2cccd2eView commit details

      Copy full SHA for 2cccd2e

      View code at this pointBrowse repository at this point
12. Commits on Apr 24, 2026

    • feat(speculative): introduce O(1) hash-based N-Gram speculative decoding

      Show description for ddf577d
      JamePeng

      committed
      ddf577dView commit details

      Copy full SHA for ddf577d

      View code at this pointBrowse repository at this point
13. Commits on Apr 22, 2026

    • feat(types): introduce MCP definitions and align with latest OpenAI spec

      Show description for afd038f
      JamePeng

      committed
      afd038fView commit details

      Copy full SHA for afd038f

      View code at this pointBrowse repository at this point
14. Commits on Apr 21, 2026

    • Update Submodule vendor/llama.cpp 81df3f7..82209ef

      JamePeng

      committed
      8625836View commit details

      Copy full SHA for 8625836

      View code at this pointBrowse repository at this point
15. Commits on Apr 20, 2026

    • Update Submodule vendor/llama.cpp 037bfe3..81df3f7

      JamePeng

      committed
      e3b7ad6View commit details

      Copy full SHA for e3b7ad6

      View code at this pointBrowse repository at this point
16. Commits on Apr 17, 2026

    • Update Submodule vendor/llama.cpp 9db77a0..45cac7c

      JamePeng

      committed
      7a19575View commit details

      Copy full SHA for 7a19575

      View code at this pointBrowse repository at this point
17. Commits on Apr 16, 2026

    • Bump version to 0.3.36

      Show description for b97cb63
      JamePeng

      committed
      b97cb63View commit details

      Copy full SHA for b97cb63

      View code at this pointBrowse repository at this point

    • feat: enhance Qwen35ChatHandler with preserve_thinking and Qwen3.6 template fixes

      Show description for 7820677
      JamePeng

      committed
      7820677View commit details

      Copy full SHA for 7820677

      View code at this pointBrowse repository at this point
18. Commits on Apr 15, 2026

    • Sync mtmd: add mtmd_image_tokens_get_decoder_pos() API (#21851)

      Show description for 9d2b2cb
      JamePeng

      committed
      9d2b2cbView commit details

      Copy full SHA for 9d2b2cb

      View code at this pointBrowse repository at this point
19. Commits on Apr 13, 2026

    • docs: add audio processing recommendation to Gemma4ChatHandler

      Show description for 5e6529e
      JamePeng

      committed
      5e6529eView commit details

      Copy full SHA for 5e6529e

      View code at this pointBrowse repository at this point
20. Commits on Apr 11, 2026

    • refactor: update Gemma4ChatHandler with latest google/gemma-4-31B-it chat template from huggingface

      Show description for 122d8db
      JamePeng

      committed
      122d8dbView commit details

      Copy full SHA for 122d8db

      View code at this pointBrowse repository at this point
21. Commits on Apr 9, 2026

    • fix missing change of llama_cpp.llama_split_mode.LLAMA_SPLIT_MODE_LAYER,

      JamePeng

      committed
      f22b819View commit details

      Copy full SHA for f22b819

      View code at this pointBrowse repository at this point