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

ggml : update mul_mat_id to use the same tensor for all the experts #6387

Merged
merged 36 commits into from
Apr 3, 2024
Merged

ggml : update mul_mat_id to use the same tensor for all the experts #6387

merged 36 commits into from
Apr 3, 2024

Commits on Mar 29, 2024

  1. ggml : update mul_mat_id to use the same tensor for all the experts

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    0c7e21d View commit details
    Browse the repository at this point in the history

  2. update cuda

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    9c9fe60 View commit details
    Browse the repository at this point in the history

  3. minor

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    2479900 View commit details
    Browse the repository at this point in the history

  4. update metal

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    93db37e View commit details
    Browse the repository at this point in the history

  5. update test-backend-ops

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    325e5ef View commit details
    Browse the repository at this point in the history

  6. fix cuda

    @slaren
    slaren committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    26c09ad View commit details
    Browse the repository at this point in the history

Commits on Mar 30, 2024

  1. Update ggml-metal.m 

    Co-authored-by: Georgi Gerganov <[email protected]>
    @slaren @ggerganov
    slaren and ggerganov committed Mar 30, 2024
    Configuration menu
    Copy the full SHA
    2abb6c7 View commit details
    Browse the repository at this point in the history

  2. update convert.py

    @slaren
    slaren committed Mar 30, 2024
    Configuration menu
    Copy the full SHA
    6203d72 View commit details
    Browse the repository at this point in the history

Commits on Mar 31, 2024

  1. update convert-hf-to-gguf.py

    @slaren
    slaren committed Mar 31, 2024
    Configuration menu
    Copy the full SHA
    4a5d50e View commit details
    Browse the repository at this point in the history

  2. update convert.py for mixtral hf models

    @slaren
    slaren committed Mar 31, 2024
    Configuration menu
    Copy the full SHA
    3b3298a View commit details
    Browse the repository at this point in the history

  3. Update convert-hf-to-gguf.py 

    Co-authored-by: Georgi Gerganov <[email protected]>
    @slaren @ggerganov
    slaren and ggerganov committed Mar 31, 2024
    Configuration menu
    Copy the full SHA
    8c2f7b8 View commit details
    Browse the repository at this point in the history

Commits on Apr 1, 2024

  1. cuda : support non-pow-2 number of experts

    @slaren
    slaren committed Apr 1, 2024
    Configuration menu
    Copy the full SHA
    4531b02 View commit details
    Browse the repository at this point in the history

  2. allow quantize to work for split and merged experts models in the sam… 

    …e way
    @slaren
    slaren committed Apr 1, 2024
    Configuration menu
    Copy the full SHA
    6886fdb View commit details
    Browse the repository at this point in the history

  3. cleanup + disable mmap automatically with split tensors models

    @slaren
    slaren committed Apr 1, 2024
    Configuration menu
    Copy the full SHA
    deea200 View commit details
    Browse the repository at this point in the history

Commits on Apr 2, 2024

  1. update imatrix

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    b4a6206 View commit details
    Browse the repository at this point in the history

  2. test-backend-ops : test qwen argsort

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    8f84ca3 View commit details
    Browse the repository at this point in the history

  3. update grok model loading

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    5de4a5d View commit details
    Browse the repository at this point in the history

  4. llama : add merged experts tensors to the grok tensor map

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    6875369 View commit details
    Browse the repository at this point in the history

  5. minor

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    6f33852 View commit details
    Browse the repository at this point in the history

  6. gguf : bump version

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    68d21de View commit details
    Browse the repository at this point in the history

  7. fix quantizing of merged experts

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    f27cbf3 View commit details
    Browse the repository at this point in the history

  8. convert-hf-to-gguf.py : update grok (untested)

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    d08a1f4 View commit details
    Browse the repository at this point in the history

  9. make linter happy

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    9530398 View commit details
    Browse the repository at this point in the history

  10. cuda/argsort : use shared memory instead of pool memory

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    f421b32 View commit details
    Browse the repository at this point in the history

  11. convert : fix grok tensor names

    @ggerganov
    ggerganov committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    c704c77 View commit details
    Browse the repository at this point in the history

  12. metal : add support for non-pow-2 argsort

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    fe62909 View commit details
    Browse the repository at this point in the history

  13. llama : more loader cleanup, better error checking

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    31adc93 View commit details
    Browse the repository at this point in the history

  14. cuda : fix warning

    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    86f3666 View commit details
    Browse the repository at this point in the history

  15. llama : still use mmap for loading old models, but copy the data to a… 

    … host buffer
    @slaren
    slaren committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    a1343ae View commit details
    Browse the repository at this point in the history

Commits on Apr 3, 2024

  1. add review note

    @slaren
    slaren committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    19dafaf View commit details
    Browse the repository at this point in the history

  2. llama : remove ffn tensor counting + add sanity check 

    ggml-ci
    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    3779b98 View commit details
    Browse the repository at this point in the history

  3. convert : fix handling of n_experts == None 

    ggml-ci
    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    e810899 View commit details
    Browse the repository at this point in the history

  4. imatrix : fix ncall counters

    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    fc719b6 View commit details
    Browse the repository at this point in the history

  5. llama : produce error if imatrix size does not match

    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    822caa4 View commit details
    Browse the repository at this point in the history

  6. quantize : terminate on errors + trace logs 

    ggml-ci
    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    a054283 View commit details
    Browse the repository at this point in the history

  7. metal : pad shared memory to 16 bytes

    @ggerganov
    ggerganov committed Apr 3, 2024
    Configuration menu
    Copy the full SHA
    716e960 View commit details
    Browse the repository at this point in the history