Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

llm : add Falcon support #2717

Merged
merged 38 commits into from
Aug 23, 2023
Merged

llm : add Falcon support #2717

merged 38 commits into from
Aug 23, 2023

Commits on Aug 22, 2023

  1. llama : refactor GGUF constants into static maps

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    4ed3469 View commit details
    Browse the repository at this point in the history

  2. llama : check if model architecture is known

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    8bd7f06 View commit details
    Browse the repository at this point in the history

  3. llama : refactor llama_model_load_internal()

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    3057d6a View commit details
    Browse the repository at this point in the history

  4. gguf : add KV constant maps

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    3c025a6 View commit details
    Browse the repository at this point in the history

  5. Merge branch 'master' into falcon

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    b19c6e4 View commit details
    Browse the repository at this point in the history

  6. llm : read arch-specific KVs

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    9f28f73 View commit details
    Browse the repository at this point in the history

  7. convert : add dummy scores + types

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    d1b3b95 View commit details
    Browse the repository at this point in the history

  8. falcon : load tensor data (CPU only)

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    2f3c80a View commit details
    Browse the repository at this point in the history

  9. llama : fix loading progress bar

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    5c5413d View commit details
    Browse the repository at this point in the history

  10. llama : add arch member to llama_model

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    085228e View commit details
    Browse the repository at this point in the history

  11. falcon : CPU inference working

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    3c7c325 View commit details
    Browse the repository at this point in the history

  12. falcon : support non-40B models

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    2d58444 View commit details
    Browse the repository at this point in the history

  13. falcon : minor

    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    0ec27ad View commit details
    Browse the repository at this point in the history

  14. llama : minor updates 

    ggml-ci
    @ggerganov
    ggerganov committed Aug 22, 2023
    Configuration menu
    Copy the full SHA
    7bbbf38 View commit details
    Browse the repository at this point in the history

  15. convert-falcon-hf-to-gguf.py : fix special token mapping

    @klosax
    klosax authored Aug 22, 2023
    Configuration menu
    Copy the full SHA
    9853f2c View commit details
    Browse the repository at this point in the history

  16. llama.cpp : llama default UNK token = id 0

    @klosax
    klosax authored Aug 22, 2023
    Configuration menu
    Copy the full SHA
    ffa5099 View commit details
    Browse the repository at this point in the history

  17. llama.cpp : fix bpe tokenizer

    @klosax
    klosax authored Aug 22, 2023
    Configuration menu
    Copy the full SHA
    a95ae75 View commit details
    Browse the repository at this point in the history

  18. llama.cpp : fix the fix of bpe tokenizer

    @klosax
    klosax authored Aug 22, 2023
    Configuration menu
    Copy the full SHA
    d561b7f View commit details
    Browse the repository at this point in the history

Commits on Aug 23, 2023

  1. ggml : pass eps to ggml_norm

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    e3c52bd View commit details
    Browse the repository at this point in the history

  2. metal : implement RoPE (mode = 2) + avoid ggml_repeat

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    99bb260 View commit details
    Browse the repository at this point in the history

  3. ggml : ggml_repeat always creates new tensor

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    af4bbcc View commit details
    Browse the repository at this point in the history

  4. falcon : copy-paste self-attention from LLaMA

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    b34ab74 View commit details
    Browse the repository at this point in the history

  5. metal : print extra compute pipeline info

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    a0dc47a View commit details
    Browse the repository at this point in the history

  6. falcon : minor changes (still chasing the Metal problem)

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    e2d23be View commit details
    Browse the repository at this point in the history

  7. llama.cpp : fix linefeed token

    @klosax
    klosax authored Aug 23, 2023
    Configuration menu
    Copy the full SHA
    b693000 View commit details
    Browse the repository at this point in the history

  8. metal : fix GELU kernel numerical stability by using precise::tanh

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    0a85ae7 View commit details
    Browse the repository at this point in the history

  9. metal : temporary workaround for the concurrency optimization bug

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    854ae5d View commit details
    Browse the repository at this point in the history

  10. falcon : add CUDA offloading (#2739)

    @slaren
    slaren authored Aug 23, 2023
    Configuration menu
    Copy the full SHA
    e729965 View commit details
    Browse the repository at this point in the history

  11. llama : better model naming and size reporting

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    176ea71 View commit details
    Browse the repository at this point in the history

  12. Merge branch 'master' into falcon

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    6938c5f View commit details
    Browse the repository at this point in the history

  13. llama : prep new tokenizer support

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    c3f8a6e View commit details
    Browse the repository at this point in the history

  14. llama : advanced BPE tokenizer based on ggllm.cpp imlpementation

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    3bfb720 View commit details
    Browse the repository at this point in the history

  15. llama : remove oboslete comment 

    ggml-ci
    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    2424e1d View commit details
    Browse the repository at this point in the history

  16. common : remove obsolete BPE API + disable test-tokenizer-1

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    596e109 View commit details
    Browse the repository at this point in the history

  17. llama : revert BPE special-case in llama_byte_to_token()

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    f8ee54b View commit details
    Browse the repository at this point in the history

  18. cuda : add TODOs for RoPE NeoX implementation

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    8c6d393 View commit details
    Browse the repository at this point in the history

  19. llama : default special tokens based on vocab type

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    630d8b4 View commit details
    Browse the repository at this point in the history

  20. perplexity : add log for start of tokenization

    @ggerganov
    ggerganov committed Aug 23, 2023
    Configuration menu
    Copy the full SHA
    fae8faa View commit details
    Browse the repository at this point in the history