Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit History

Commits on Feb 26, 2026

• feat(core): overhaul generate and eval for hybrid model support(Qwen3-next、Qwen3.5 etc.)

  Show description for 978ddf7
  JamePeng

  committed
  978ddf7

  Copy full SHA for 978ddf7

• refactor(LlamaBatch): replace set_batch with granular add_token + vectorized add_sequence

  Show description for 9262fc0
  JamePeng

  committed
  9262fc0

  Copy full SHA for 9262fc0

Commits on Feb 13, 2026

• test: replace unstable grammar test with deterministic mechanism check

  Show description for a0e3ab4
  JamePeng

  committed
  a0e3ab4

  Copy full SHA for a0e3ab4

Commits on Feb 9, 2026

• test: update integration tests for new sampler architecture

  Show description for 31a9d73
  JamePeng

  committed
  31a9d73

  Copy full SHA for 31a9d73

• Refactor Llama class to use new LlamaSampler chain API from _internals

  Show description for 1e6094a
  JamePeng

  committed
  1e6094a

  Copy full SHA for 1e6094a

Commits on Jan 29, 2026

• Sync llama : disable Direct IO by default

  JamePeng

  committed
  5bdff0b

  Copy full SHA for 5bdff0b

Commits on Jan 11, 2026

• Fixed test_llama.py

  JamePeng

  committed
  6d3cfc3

  Copy full SHA for 6d3cfc3

Commits on Jan 8, 2026

• Update llama.cpp API 20260108

  JamePeng

  committed
  4d0fe24

  Copy full SHA for 4d0fe24

Commits on Jan 1, 2026

• Fixed `embeddings` typos

  JamePeng

  committed
  8a96113

  Copy full SHA for 8a96113

Commits on Sep 13, 2025

• Sync aLoRA Support

  JamePeng

  committed
  3d83305

  Copy full SHA for 3d83305

Commits on Aug 31, 2025

• Sync llama: use FA + max. GPU layers by default

  JamePeng

  committed
  2be720d

  Copy full SHA for 2be720d

Commits on Jul 16, 2025

• Update llama.cpp API 20250717

  JamePeng

  committed
  bc283d3

  Copy full SHA for bc283d3

Commits on Jul 14, 2025

• try to use the logit_bias instead of logit_processors in test_llama

  JamePeng

  committed
  790cc09

  Copy full SHA for 790cc09

Commits on Jul 6, 2025

• fix missing logits_all params

  JamePeng

  committed
  51e77fa

  Copy full SHA for 51e77fa

• Sync abetlen/llama-cpp-python 0.3.11 code

  JamePeng

  committed
  0701c73

  Copy full SHA for 0701c73

Commits on May 20, 2025

• Sync kv-cache : add SWA support

  JamePeng

  committed
  177445c

  Copy full SHA for 177445c

Commits on May 9, 2025

• Sync context : remove logits_all flag and update API

  JamePeng

  committed
  1a1d920

  Copy full SHA for 1a1d920

Commits on Sep 19, 2024

• fix: Fix memory allocation of ndarray (#1704)

  Show description for 22cedad
  

  xu-song

  and

  abetlen

  authored
  22cedad

  Copy full SHA for 22cedad

• feat: Update sampling API for llama.cpp (#1742)

  Show description for f8fcb3e
  abetlen

  authored
  f8fcb3e

  Copy full SHA for f8fcb3e

Commits on Jun 4, 2024

• fix: Avoid duplicate special tokens in chat formats (#1439)

  Show description for 027f7bc
  

  CISC

  and

  abetlen

  authored
  027f7bc

  Copy full SHA for 027f7bc

Commits on Apr 30, 2024

• feat: Update llama.cpp

  abetlen

  committed
  97fb860

  Copy full SHA for 97fb860

Commits on Feb 26, 2024

• feat: Update llama.cpp

  abetlen

  committed
  cbbcd88

  Copy full SHA for cbbcd88

Commits on Feb 21, 2024

• feat(low-level-api): Improve API static type-safety and performance (#1205)

  abetlen

  authored
  7f51b60

  Copy full SHA for 7f51b60

Commits on Feb 8, 2024

• misc: rename grammar test

  abetlen

  committed
  2ef7ba3

  Copy full SHA for 2ef7ba3

Commits on Jan 31, 2024

• Add speculative decoding (#1120)

  Show description for fb762a6
  abetlen

  authored
  fb762a6

  Copy full SHA for fb762a6

Commits on Jan 29, 2024

• Add chat format test.

  abetlen

  committed
  9ae5819

  Copy full SHA for 9ae5819

Commits on Jan 22, 2024

• fix: from_json_schema oneof/anyof bug. Closes #1097

  abetlen

  committed
  d3f5528

  Copy full SHA for d3f5528

Commits on Jan 19, 2024

• feat: Add ability to load chat format from huggingface autotokenizer or tokenizer_config.json files.

  abetlen

  committed
  b8fc1c7

  Copy full SHA for b8fc1c7

Commits on Jan 17, 2024

• Integration of Jinja2 Templating (#875)

  Show description for 6bfe98b
  teleprint-me

  authored
  6bfe98b

  Copy full SHA for 6bfe98b

Commits on Jan 15, 2024

• Fix Pydantic model parsing (#1087)

  DeNeutoy

  authored
  c689ccc

  Copy full SHA for c689ccc

Commits on Dec 16, 2023

• Fix logits_to_logprobs for 2-D and 3-D logits (#1002)

  Show description for 5a89446
  kddubey

  authored
  5a89446

  Copy full SHA for 5a89446

Commits on Nov 22, 2023

• tests: add mock_kv_cache placeholder functions

  abetlen

  committed
  9515467

  Copy full SHA for 9515467

• tests: avoid constantly reallocating logits

  abetlen

  committed
  0ea2444

  Copy full SHA for 0ea2444

• tests: don't mock sampling functions

  abetlen

  committed
  0a7e05b

  Copy full SHA for 0a7e05b

Commits on Nov 21, 2023

• Use mock_llama for all tests

  abetlen

  committed
  d7388f1

  Copy full SHA for d7388f1