Commits

Commit history

1. Commits on Sep 20, 2024

   • feat: Add option to configure n_ubatch

     abetlen

     committed
     6c44a3fView commit details

     Copy full SHA for 6c44a3f

     View code at this pointBrowse repository at this point
2. Commits on Sep 19, 2024

   • feat: Add loading sharded GGUF files from HuggingFace with Llama.from_pretrained(additional_files=[...]) . Closes #1341

     Show description for 84c0920
     

     Gnurro

     and

     abetlen

     authored
     84c0920View commit details

     Copy full SHA for 84c0920

     View code at this pointBrowse repository at this point

   • fix: Don't store scores internally unless logits_all=True. Reduces memory requirements for large context. Closes #1542

     abetlen

     committed
     29afcfdView commit details

     Copy full SHA for 29afcfd

     View code at this pointBrowse repository at this point

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

     Show description for 22cedad
     

     xu-song

     and

     abetlen

     authored
     22cedadView commit details

     Copy full SHA for 22cedad

     View code at this pointBrowse repository at this point

   • misc: Format

     abetlen

     committed
     9b64bb5View commit details

     Copy full SHA for 9b64bb5

     View code at this pointBrowse repository at this point

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

     Show description for f8fcb3e
     abetlen

     authored
     f8fcb3eView commit details

     Copy full SHA for f8fcb3e

     View code at this pointBrowse repository at this point
3. Commits on Aug 31, 2024

   • feat: Update llama.cpp

     abetlen

     committed
     9769e57View commit details

     Copy full SHA for 9769e57

     View code at this pointBrowse repository at this point
4. Commits on Aug 29, 2024

   • feat: Enable detokenizing special tokens with `special=True` (#1596)

     Show description for d981d32
     

     benniekiss

     and

     abetlen

     authored
     d981d32View commit details

     Copy full SHA for d981d32

     View code at this pointBrowse repository at this point
5. Commits on Aug 15, 2024

   • fix: Llama.close didn't free lora adapter (#1679)

     jkawamoto

     authored
     3c7501bView commit details

     Copy full SHA for 3c7501b

     View code at this pointBrowse repository at this point

   • feat: Update llama.cpp

     abetlen

     committed
     63d65acView commit details

     Copy full SHA for 63d65ac

     View code at this pointBrowse repository at this point
6. Commits on Aug 12, 2024

   • fix: only print 'cache saved' in verbose mode (#1668)

     Show description for 9bab46f
     

     lsorber

     and

     abetlen

     authored
     9bab46fView commit details

     Copy full SHA for 9bab46f

     View code at this pointBrowse repository at this point
7. Commits on Aug 8, 2024

   • fix: typo

     abetlen

     committed
     7aaf701View commit details

     Copy full SHA for 7aaf701

     View code at this pointBrowse repository at this point
8. Commits on Aug 7, 2024

   • feat: Enable recursive search of HFFS.ls when using `from_pretrained` (#1656)

     Show description for 5e39a85
     

     benHeid

     and

     abetlen

     authored
     5e39a85View commit details

     Copy full SHA for 5e39a85

     View code at this pointBrowse repository at this point

   • feat: Add more detailed log for prefix-match (#1659)

     Show description for e966f3b
     

     xu-song

     and

     abetlen

     authored
     e966f3bView commit details

     Copy full SHA for e966f3b

     View code at this pointBrowse repository at this point
9. Commits on Jul 31, 2024

   • fix : Missing LoRA adapter after API change (#1630)

     shamitv

     authored
     1f0b9a2View commit details

     Copy full SHA for 1f0b9a2

     View code at this pointBrowse repository at this point
10. Commits on Jul 18, 2024

    • fix: Change repeat_penalty to 1.0 to match llama.cpp defaults (#1590)

      Show description for 0700476
      

      ddh0

      and

      abetlen

      authored
      0700476View commit details

      Copy full SHA for 0700476

      View code at this pointBrowse repository at this point
11. Commits on Jul 17, 2024

    • feat: Update llama.cpp

      abetlen

      committed
      7613d23View commit details

      Copy full SHA for 7613d23

      View code at this pointBrowse repository at this point
12. Commits on Jul 9, 2024

    • fix(minor): Minor ruff fixes

      abetlen

      committed
      08f2bb3View commit details

      Copy full SHA for 08f2bb3

      View code at this pointBrowse repository at this point

    • fix(misc): Format

      abetlen

      committed
      c1ae815View commit details

      Copy full SHA for c1ae815

      View code at this pointBrowse repository at this point
13. Commits on Jun 19, 2024

    • fix: Make destructor to automatically call .close() method on Llama class.

      abetlen

      committed
      4c1d74cView commit details

      Copy full SHA for 4c1d74c

      View code at this pointBrowse repository at this point
14. Commits on Jun 13, 2024

    • feat: Add `.close()` method to `Llama` class to explicitly free model from memory (#1513)

      Show description for 320a5d7
      

      jkawamoto

      and

      abetlen

      authored
      320a5d7View commit details

      Copy full SHA for 320a5d7

      View code at this pointBrowse repository at this point

    • feat: Support SPM infill (#1492)

      Show description for dbcf64c
      

      CISC

      and

      abetlen

      authored
      dbcf64cView commit details

      Copy full SHA for dbcf64c

      View code at this pointBrowse repository at this point
15. Commits on Jun 4, 2024

    • feat: adding `rpc_servers` parameter to `Llama` class (#1477)

      Show description for d634efc
      

      chraac

      and

      abetlen

      authored
      d634efcView commit details

      Copy full SHA for d634efc

      View code at this pointBrowse repository at this point

    • fix: fix logprobs when BOS is not present (#1471)

      Show description for 6e0642c
      a-ghorbani

      authored
      6e0642cView commit details

      Copy full SHA for 6e0642c

      View code at this pointBrowse repository at this point

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

      Show description for 027f7bc
      

      CISC

      and

      abetlen

      authored
      027f7bcView commit details

      Copy full SHA for 027f7bc

      View code at this pointBrowse repository at this point
16. Commits on May 29, 2024

    • fix: fix string value kv_overrides. Closes #1487

      abetlen

      committed
      df45a4bView commit details

      Copy full SHA for df45a4b

      View code at this pointBrowse repository at this point
17. Commits on May 24, 2024

    • feat: Improve Llama.eval performance by avoiding list conversion (#1476)

      Show description for 5cae104
      

      thoughtp0lice

      and

      abetlen

      authored
      5cae104View commit details

      Copy full SHA for 5cae104

      View code at this pointBrowse repository at this point
18. Commits on May 16, 2024

    • fix: segfault for models without eos / bos tokens. Closes #1463

      abetlen

      committed
      d99a6baView commit details

      Copy full SHA for d99a6ba

      View code at this pointBrowse repository at this point
19. Commits on May 14, 2024

    • feat: add MinTokensLogitProcessor and min_tokens argument to server (#1333)

      Show description for 5212fb0
      twaka

      authored
      5212fb0View commit details

      Copy full SHA for 5212fb0

      View code at this pointBrowse repository at this point

    • misc: Remove unnecessary metadata lookups (#1448)

      Show description for 389e09c
      CISC

      authored
      389e09cView commit details

      Copy full SHA for 389e09c

      View code at this pointBrowse repository at this point
20. Commits on May 9, 2024

    • feat: Support multiple chat templates - step 1 (#1396)

      Show description for 5ab40e6
      

      CISC

      and

      abetlen

      authored
      5ab40e6View commit details

      Copy full SHA for 5ab40e6

      View code at this pointBrowse repository at this point
21. Commits on May 8, 2024

    • feat: fill-in-middle support (#1386)

      Show description for 4a7122d
      

      CISC

      and

      abetlen

      authored
      4a7122dView commit details

      Copy full SHA for 4a7122d

      View code at this pointBrowse repository at this point

    • fix: chat_format log where auto-detected format prints `None` (#1434)

      balvisio

      authored
      a50d24eView commit details

      Copy full SHA for a50d24e

      View code at this pointBrowse repository at this point
22. Commits on May 3, 2024

    • fix: Use memmove to copy str_value kv_override. Closes #1417

      abetlen

      committed
      9f7a855View commit details

      Copy full SHA for 9f7a855

      View code at this pointBrowse repository at this point
23. Commits on Apr 30, 2024

    • fix: wrong parameter for flash attention in pickle __getstate__

      abetlen

      committed
      29b6e9aView commit details

      Copy full SHA for 29b6e9a

      View code at this pointBrowse repository at this point