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] Support fuse attention q, k, v weights #8202

Merged
merged 29 commits into from
Apr 25, 2024
Merged

[LLM] Support fuse attention q, k, v weights #8202

merged 29 commits into from
Apr 25, 2024

Commits on Mar 18, 2024

  1. 1. add use-interface & fuse action 

    1.1. modify 1., code order

2. switch to name_mapping

3. solve tp branch

3.2 follow hui, handel qkv separately

3.3 handle pdparams

3.4 from torch

3.5 abandon low_cpu_mem_usage

3.6 solve shard branch
    @ziangqin-baidu
    ziangqin-baidu committed Mar 18, 2024
    Configuration menu
    Copy the full SHA
    d859826 View commit details
    Browse the repository at this point in the history

  2. 3.6.1 solve shard branch after rebase develop

    @ziangqin-baidu
    ziangqin-baidu committed Mar 18, 2024
    Configuration menu
    Copy the full SHA
    f7b6973 View commit details
    Browse the repository at this point in the history

Commits on Mar 19, 2024

  1. code clean

    @ziangqin-baidu
    ziangqin-baidu committed Mar 19, 2024
    Configuration menu
    Copy the full SHA
    05507a7 View commit details
    Browse the repository at this point in the history

Commits on Mar 28, 2024

  1. Merge remote-tracking branch 'qinziang/fuse' into dev-fuse-qkv

    @DrownFish19
    DrownFish19 committed Mar 28, 2024
    Configuration menu
    Copy the full SHA
    b8b828b View commit details
    Browse the repository at this point in the history

Commits on Mar 29, 2024

  1. remove debug comment

    @DrownFish19
    DrownFish19 committed Mar 29, 2024
    Configuration menu
    Copy the full SHA
    28ed30f View commit details
    Browse the repository at this point in the history

Commits on Apr 1, 2024

  1. Redefine fuse and split functions

    @DrownFish19
    DrownFish19 committed Apr 1, 2024
    Configuration menu
    Copy the full SHA
    1a1591a View commit details
    Browse the repository at this point in the history

  2. Redefine fuse and split functions

    @DrownFish19
    DrownFish19 committed Apr 1, 2024
    Configuration menu
    Copy the full SHA
    ef1fb18 View commit details
    Browse the repository at this point in the history

Commits on Apr 2, 2024

  1. Merge branch 'dev-fuse-qkv' of github.com:DrownFish19/PaddleNLP into … 

    …dev-fuse-qkv
    @DrownFish19
    DrownFish19 committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    f56e602 View commit details
    Browse the repository at this point in the history

  2. comment and fix

    @DrownFish19
    DrownFish19 committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    19faa26 View commit details
    Browse the repository at this point in the history

  3. update method

    @DrownFish19
    DrownFish19 committed Apr 2, 2024
    Configuration menu
    Copy the full SHA
    4a92dde View commit details
    Browse the repository at this point in the history

Commits on Apr 9, 2024

  1. update QKV fuse and split

    @DrownFish19
    DrownFish19 committed Apr 9, 2024
    Configuration menu
    Copy the full SHA
    c0a71d2 View commit details
    Browse the repository at this point in the history

Commits on Apr 10, 2024

  1. support fuse weights in multi-files

    @DrownFish19
    DrownFish19 committed Apr 10, 2024
    Configuration menu
    Copy the full SHA
    d1e7d17 View commit details
    Browse the repository at this point in the history

  2. add precision compare

    @DrownFish19
    DrownFish19 committed Apr 10, 2024
    Configuration menu
    Copy the full SHA
    58a2c23 View commit details
    Browse the repository at this point in the history

Commits on Apr 11, 2024

  1. simplify function call

    @DrownFish19
    DrownFish19 committed Apr 11, 2024
    Configuration menu
    Copy the full SHA
    7e1ab08 View commit details
    Browse the repository at this point in the history

  2. support use_fast_ffn

    @DrownFish19
    DrownFish19 committed Apr 11, 2024
    Configuration menu
    Copy the full SHA
    0458f4e View commit details
    Browse the repository at this point in the history

  3. clean modeling and configuration

    @DrownFish19
    DrownFish19 committed Apr 11, 2024
    Configuration menu
    Copy the full SHA
    920155b View commit details
    Browse the repository at this point in the history

  4. add test for gpt and opt

    @DrownFish19
    DrownFish19 committed Apr 11, 2024
    Configuration menu
    Copy the full SHA
    a90fcfc View commit details
    Browse the repository at this point in the history

  5. fix tp_actions get

    @DrownFish19
    DrownFish19 committed Apr 11, 2024
    Configuration menu
    Copy the full SHA
    774bb6f View commit details
    Browse the repository at this point in the history

  6. Merge branch 'PaddlePaddle:develop' into dev-fuse-qkv

    @DrownFish19
    DrownFish19 authored Apr 11, 2024
    Configuration menu
    Copy the full SHA
    057198c View commit details
    Browse the repository at this point in the history

Commits on Apr 12, 2024

  1. Merge branch 'PaddlePaddle:develop' into dev-fuse-qkv

    @DrownFish19
    DrownFish19 authored Apr 12, 2024
    Configuration menu
    Copy the full SHA
    db78c6a View commit details
    Browse the repository at this point in the history

  2. Merge branch 'PaddlePaddle:develop' into dev-fuse-qkv

    @DrownFish19
    DrownFish19 authored Apr 12, 2024
    Configuration menu
    Copy the full SHA
    a1aa078 View commit details
    Browse the repository at this point in the history

  3. add fast_ffn test

    @DrownFish19
    DrownFish19 committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    f0d19f6 View commit details
    Browse the repository at this point in the history

  4. Merge branch 'dev-fuse-qkv' of github.com:DrownFish19/PaddleNLP into … 

    …dev-fuse-qkv
    @DrownFish19
    DrownFish19 committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    110983d View commit details
    Browse the repository at this point in the history

Commits on Apr 16, 2024

  1. add Qwen2Moe

    @DrownFish19
    DrownFish19 committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    113b883 View commit details
    Browse the repository at this point in the history

Commits on Apr 17, 2024

  1. Revert "add Qwen2Moe" 

    This reverts commit 113b883.
    @DrownFish19
    DrownFish19 committed Apr 17, 2024
    Configuration menu
    Copy the full SHA
    1fb0a7d View commit details
    Browse the repository at this point in the history

  2. Merge branch 'PaddlePaddle:develop' into dev-fuse-qkv

    @DrownFish19
    DrownFish19 authored Apr 17, 2024
    Configuration menu
    Copy the full SHA
    a118f1b View commit details
    Browse the repository at this point in the history

  3. add test for split

    @DrownFish19
    DrownFish19 committed Apr 17, 2024
    Configuration menu
    Copy the full SHA
    11d0793 View commit details
    Browse the repository at this point in the history

  4. update doc

    @DrownFish19
    DrownFish19 committed Apr 17, 2024
    Configuration menu
    Copy the full SHA
    0c98a45 View commit details
    Browse the repository at this point in the history

Commits on Apr 18, 2024

  1. update filter_dict_keys

    @DrownFish19
    DrownFish19 committed Apr 18, 2024
    Configuration menu
    Copy the full SHA
    f6f3b0e View commit details
    Browse the repository at this point in the history