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

Fix Sym=False, new checkpoint_format = gptq_v2 #9

Merged
merged 69 commits into from
Jun 15, 2024
Merged

Fix Sym=False, new checkpoint_format = gptq_v2 #9

merged 69 commits into from
Jun 15, 2024

Commits on Feb 21, 2024

  1. remove (zeors -= 1)

    @qwopqwop200
    qwopqwop200 committed Feb 21, 2024
    Configuration menu
    Copy the full SHA
    b33e1d0 View commit details
    Browse the repository at this point in the history

  2. add warning

    @qwopqwop200
    qwopqwop200 committed Feb 21, 2024
    Configuration menu
    Copy the full SHA
    6419a2d View commit details
    Browse the repository at this point in the history

Commits on Feb 22, 2024

  1. support backwards compatibility

    @qwopqwop200
    qwopqwop200 committed Feb 22, 2024
    Configuration menu
    Copy the full SHA
    5ea98bb View commit details
    Browse the repository at this point in the history

Commits on Feb 25, 2024

  1. support and fix bug

    @qwopqwop200
    qwopqwop200 committed Feb 25, 2024
    Configuration menu
    Copy the full SHA
    0b07292 View commit details
    Browse the repository at this point in the history

Commits on Feb 26, 2024

  1. remove not necessary parm

    @qwopqwop200
    qwopqwop200 authored Feb 26, 2024
    Configuration menu
    Copy the full SHA
    b015ae9 View commit details
    Browse the repository at this point in the history

Commits on Mar 2, 2024

  1. fix test_q4 bug

    @qwopqwop200
    qwopqwop200 committed Mar 2, 2024
    Configuration menu
    Copy the full SHA
    b7d9ade View commit details
    Browse the repository at this point in the history

  2. fix test_q4 bug

    @qwopqwop200
    qwopqwop200 committed Mar 2, 2024
    Configuration menu
    Copy the full SHA
    83e510e View commit details
    Browse the repository at this point in the history

  3. fix bug double converting

    @qwopqwop200
    qwopqwop200 committed Mar 2, 2024
    Configuration menu
    Copy the full SHA
    a89cb77 View commit details
    Browse the repository at this point in the history

  4. Update _utils.py

    @qwopqwop200
    qwopqwop200 authored Mar 2, 2024
    Configuration menu
    Copy the full SHA
    639e66a View commit details
    Browse the repository at this point in the history

  5. Merge branch 'main' into main

    @qwopqwop200
    qwopqwop200 authored Mar 2, 2024
    Configuration menu
    Copy the full SHA
    15ecb0f View commit details
    Browse the repository at this point in the history

Commits on Apr 12, 2024

  1. merge main #1

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    0759aea View commit details
    Browse the repository at this point in the history

  2. FIX type error

    @LRL-ModelCloud
    LRL-ModelCloud committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    101647d View commit details
    Browse the repository at this point in the history

  3. module is nn.Module

    @LRL-ModelCloud
    LRL-ModelCloud committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    28a2541 View commit details
    Browse the repository at this point in the history

  4. sync name

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    3899d5b View commit details
    Browse the repository at this point in the history

  5. need return module

    @LRL-ModelCloud
    LRL-ModelCloud committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    9f25c21 View commit details
    Browse the repository at this point in the history

  6. modify default format to gptq_v2

    @LRL-ModelCloud
    LRL-ModelCloud committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    4159474 View commit details
    Browse the repository at this point in the history

  7. fix need return model

    @LRL-ModelCloud
    LRL-ModelCloud committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    c853143 View commit details
    Browse the repository at this point in the history

  8. remove fixme and default to gptq_v2 for quantize_config

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    047dd97 View commit details
    Browse the repository at this point in the history

  9. save _qlinear_kernel and allow save to older format

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    869a162 View commit details
    Browse the repository at this point in the history

  10. fix name

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    365d961 View commit details
    Browse the repository at this point in the history

  11. pass quantize

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    e363d85 View commit details
    Browse the repository at this point in the history

  12. Merge remote-tracking branch 'origin/main' into sym-false

    @Qubitium
    Qubitium committed Apr 12, 2024
    Configuration menu
    Copy the full SHA
    6af41db View commit details
    Browse the repository at this point in the history

Commits on Apr 13, 2024

  1. update

    @Qubitium
    Qubitium committed Apr 13, 2024
    Configuration menu
    Copy the full SHA
    bd12bde View commit details
    Browse the repository at this point in the history

  2. store quant log/stats in dict slice and return to user in quantize()

    @Qubitium
    Qubitium committed Apr 13, 2024
    Configuration menu
    Copy the full SHA
    cd916fe View commit details
    Browse the repository at this point in the history

  3. accept saved quant_log in quantize() and calculate diff

    @Qubitium
    Qubitium committed Apr 13, 2024
    Configuration menu
    Copy the full SHA
    d35e57d View commit details
    Browse the repository at this point in the history

Commits on Apr 14, 2024

  1. tqdm the layer loop

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    b0d2ad5 View commit details
    Browse the repository at this point in the history

  2. log awq vs autogptq outputs in awq compat test

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    45cbc6b View commit details
    Browse the repository at this point in the history

  3. fix cached models is not compatible with new pr. add v2 to cache file… 

    … name
    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    23b4b65 View commit details
    Browse the repository at this point in the history

  4. add deprecation warning for loading .bin/.pt weights

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    dc5d3dd View commit details
    Browse the repository at this point in the history

  5. add missing termcolor req

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    fec3f1c View commit details
    Browse the repository at this point in the history

  6. spell

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    2921b27 View commit details
    Browse the repository at this point in the history

  7. fix triton v2

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    7b97f73 View commit details
    Browse the repository at this point in the history

  8. rename quant log column 'name' to 'module'

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    c5c98da View commit details
    Browse the repository at this point in the history

  9. ruff

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    5b95194 View commit details
    Browse the repository at this point in the history

  10. add quantization tests for sym=False

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    19189aa View commit details
    Browse the repository at this point in the history

  11. spell

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    070e6c8 View commit details
    Browse the repository at this point in the history

  12. fix type hint

    @Qubitium
    Qubitium committed Apr 14, 2024
    Configuration menu
    Copy the full SHA
    0103785 View commit details
    Browse the repository at this point in the history

Commits on Apr 15, 2024

  1. more testing, fix serialization bug, no additional dependency

    @fxmarty
    fxmarty committed Apr 15, 2024
    Configuration menu
    Copy the full SHA
    61a8713 View commit details
    Browse the repository at this point in the history

Commits on Apr 16, 2024

  1. fix version

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    c6b3632 View commit details
    Browse the repository at this point in the history

  2. no need for ... in tqdm

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    83e932d View commit details
    Browse the repository at this point in the history

  3. Use threadpoolctl to limit packing threads

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    a3863a4 View commit details
    Browse the repository at this point in the history

  4. layer # sync with visual tqdm

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    7b89b89 View commit details
    Browse the repository at this point in the history

  5. use thread limit 1: as good as 4 and 1 beats 16 threads in testing

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    ed0fe60 View commit details
    Browse the repository at this point in the history

  6. fix saving of gptq (v1)

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    37ed02e View commit details
    Browse the repository at this point in the history

  7. deep copy

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    5d66558 View commit details
    Browse the repository at this point in the history

  8. remove todo: verified

    @Qubitium
    Qubitium committed Apr 16, 2024
    Configuration menu
    Copy the full SHA
    9d88639 View commit details
    Browse the repository at this point in the history

Commits on Apr 18, 2024

  1. TEST/DEBUG underflow protection and output underflow stats 

    fix underflow cond reversed
    @Qubitium
    Qubitium committed Apr 18, 2024
    Configuration menu
    Copy the full SHA
    0b84da7 View commit details
    Browse the repository at this point in the history

  2. force underflow math (testing shows this is better than skipping math)

    @Qubitium
    Qubitium committed Apr 18, 2024
    Configuration menu
    Copy the full SHA
    9e3fea0 View commit details
    Browse the repository at this point in the history

Commits on Apr 22, 2024

  1. 1) disable serialization of sym=False to v1 by default. 2) disable lo… 

    …ading of v1 sym=False by default.
    @Qubitium
    Qubitium committed Apr 22, 2024
    Configuration menu
    Copy the full SHA
    005a942 View commit details
    Browse the repository at this point in the history

Commits on Apr 23, 2024

  1. revert adding underflow check/stats

    @Qubitium
    Qubitium committed Apr 23, 2024
    Configuration menu
    Copy the full SHA
    365618d View commit details
    Browse the repository at this point in the history

  2. pass test_quant and test both v1 and v2 save/load

    @Qubitium
    Qubitium committed Apr 23, 2024
    Configuration menu
    Copy the full SHA
    31d4027 View commit details
    Browse the repository at this point in the history

Commits on Apr 24, 2024

  1. performance fix for convert_v1/v2().

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    ac63033 View commit details
    Browse the repository at this point in the history

  2. need to ++ version so can delimit models make pre/post pr

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    e8eabe2 View commit details
    Browse the repository at this point in the history

  3. add meta and meta.quantizer to quantized_config.json

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    e4936f5 View commit details
    Browse the repository at this point in the history

  4. fix json save and add meta check to test_quantization. distutils is d… 

    …eprecated by python. add packaging. depend
    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    c822d63 View commit details
    Browse the repository at this point in the history

  5. fix failed test

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    35cc701 View commit details
    Browse the repository at this point in the history

  6. fix awq unpack/repacking thread regression

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    cddbe23 View commit details
    Browse the repository at this point in the history

  7. remove highly flaky mistral tiny test with input/output that is nonse… 

    …nsical
    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    811ca39 View commit details
    Browse the repository at this point in the history

  8. now we can detect quant producer, we don't need use_unsafe_math for l… 

    …oading
    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    2104cae View commit details
    Browse the repository at this point in the history

  9. updat tests

    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    bc2bf5b View commit details
    Browse the repository at this point in the history

  10. default to gptq v1 for max compat and remove use_unsafe_math check in… 

    … save_quantized
    @Qubitium
    Qubitium committed Apr 24, 2024
    Configuration menu
    Copy the full SHA
    528a8fc View commit details
    Browse the repository at this point in the history

Commits on Apr 25, 2024

  1. misc

    @Qubitium
    Qubitium committed Apr 25, 2024
    Configuration menu
    Copy the full SHA
    464fc7e View commit details
    Browse the repository at this point in the history

  2. separate the concept of meta.quantizer and meta.packer (intel/auto-ro… 

    …und as example)
    @Qubitium
    Qubitium committed Apr 25, 2024
    Configuration menu
    Copy the full SHA
    59be4b3 View commit details
    Browse the repository at this point in the history

  3. clean

    @Qubitium
    Qubitium committed Apr 25, 2024
    Configuration menu
    Copy the full SHA
    79050ea View commit details
    Browse the repository at this point in the history

  4. test allow loading quantized lm_head

    @Qubitium
    Qubitium committed Apr 25, 2024
    Configuration menu
    Copy the full SHA
    4450702 View commit details
    Browse the repository at this point in the history

  5. rename

    @Qubitium
    Qubitium committed Apr 25, 2024
    Configuration menu
    Copy the full SHA
    d85f3f2 View commit details
    Browse the repository at this point in the history

Commits on Apr 26, 2024

  1. fix quantized lm_head loading

    @Qubitium
    Qubitium committed Apr 26, 2024
    Configuration menu
    Copy the full SHA
    55336b5 View commit details
    Browse the repository at this point in the history

  2. sync with main

    @Qubitium
    Qubitium committed Apr 26, 2024
    Configuration menu
    Copy the full SHA
    d086437 View commit details
    Browse the repository at this point in the history

Commits on Jun 6, 2024

  1. ADD GLM model support

    @LRL-ModelCloud
    LRL-ModelCloud committed Jun 6, 2024
    Configuration menu
    Copy the full SHA
    fe191f5 View commit details
    Browse the repository at this point in the history