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

[WIP] SmoothQuant using tensor subclassing #1030

Draft
wants to merge 38 commits into
base: main
Choose a base branch
from
Draft

[WIP] SmoothQuant using tensor subclassing #1030

wants to merge 38 commits into from
+805 −2

Commits on Sep 30, 2024

  1. SmoothQuant using tensor subclassing

    @Xia-Weiwen
    Xia-Weiwen committed Sep 30, 2024
    Configuration menu
    Copy the full SHA
    d34859b View commit details
    Browse the repository at this point in the history

Commits on Oct 8, 2024

  1. Merge branch 'main' into smooth_quant

    @Xia-Weiwen
    Xia-Weiwen committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    df5b49f View commit details
    Browse the repository at this point in the history

  2. Update UT

    @Xia-Weiwen
    Xia-Weiwen committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    847f1f2 View commit details
    Browse the repository at this point in the history

  3. Add SmoothQuant example

    @Xia-Weiwen
    Xia-Weiwen committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    f03cfb3 View commit details
    Browse the repository at this point in the history

Commits on Oct 9, 2024

  1. Remove duplicate implementation of int_scaled_matmul for CPU

    @Xia-Weiwen
    Xia-Weiwen committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    a2518f1 View commit details
    Browse the repository at this point in the history

  2. Update example.py

    @Xia-Weiwen
    Xia-Weiwen committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    28fb8ce View commit details
    Browse the repository at this point in the history

  3. Remove unused code

    @Xia-Weiwen
    Xia-Weiwen committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    bada2b0 View commit details
    Browse the repository at this point in the history

Commits on Oct 10, 2024

  1. Implement with LinearActivationQuantizedTensor

    @Xia-Weiwen
    Xia-Weiwen committed Oct 10, 2024
    Configuration menu
    Copy the full SHA
    921efc0 View commit details
    Browse the repository at this point in the history

  2. Fix load/save

    @Xia-Weiwen
    Xia-Weiwen committed Oct 10, 2024
    Configuration menu
    Copy the full SHA
    ad5b97e View commit details
    Browse the repository at this point in the history

  3. Fix device mismatch in observer

    @Xia-Weiwen
    Xia-Weiwen committed Oct 10, 2024
    Configuration menu
    Copy the full SHA
    f1be01d View commit details
    Browse the repository at this point in the history

Commits on Oct 11, 2024

  1. Fix fp16 overflow issue in int_scaled_matmul

    @Xia-Weiwen
    Xia-Weiwen committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    7ee1f13 View commit details
    Browse the repository at this point in the history

  2. Merge branch 'main' into smooth_quant

    @Xia-Weiwen
    Xia-Weiwen committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    c773386 View commit details
    Browse the repository at this point in the history

  3. Add linear_activation_scale_quantized.py for torch.compile

    @Xia-Weiwen
    Xia-Weiwen committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    427ff73 View commit details
    Browse the repository at this point in the history

Commits on Oct 12, 2024

  1. Quantize act/wei to 7 bit on old CPU platforms

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    9916113 View commit details
    Browse the repository at this point in the history

  2. Fix device mismatch

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    52260b6 View commit details
    Browse the repository at this point in the history

  3. Fix UT failures

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    ca50fee View commit details
    Browse the repository at this point in the history

  4. Fix UT

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    3e90789 View commit details
    Browse the repository at this point in the history

  5. Don't use torch._int_mm for CPU now because it may overflow

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    d47fcc1 View commit details
    Browse the repository at this point in the history

  6. Remove reduce_range

    @Xia-Weiwen
    Xia-Weiwen committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    a195e73 View commit details
    Browse the repository at this point in the history

Commits on Oct 14, 2024

  1. Refine code

    @Xia-Weiwen
    Xia-Weiwen committed Oct 14, 2024
    Configuration menu
    Copy the full SHA
    6627be1 View commit details
    Browse the repository at this point in the history

  2. Remove torch.compile from example

    @Xia-Weiwen
    Xia-Weiwen committed Oct 14, 2024
    Configuration menu
    Copy the full SHA
    fb981e7 View commit details
    Browse the repository at this point in the history

Commits on Oct 15, 2024

  1. Add torch.compile in example

    @Xia-Weiwen
    Xia-Weiwen committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    17c374e View commit details
    Browse the repository at this point in the history

  2. Debug CI failures

    @Xia-Weiwen
    Xia-Weiwen committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    bb76de6 View commit details
    Browse the repository at this point in the history

  3. Debug CI failures (1)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    98b2de1 View commit details
    Browse the repository at this point in the history

  4. Debug CI failures (2)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    316f5ea View commit details
    Browse the repository at this point in the history

Commits on Oct 16, 2024

  1. Debug CI failures (3)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    b4d8383 View commit details
    Browse the repository at this point in the history

  2. Work with torch.compile

    @Xia-Weiwen
    Xia-Weiwen committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    aca06d2 View commit details
    Browse the repository at this point in the history

  3. Update torchao/kernel/intmm.py

    @Xia-Weiwen
    Xia-Weiwen committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    dde7545 View commit details
    Browse the repository at this point in the history

Commits on Oct 17, 2024

  1. Update readme.md

    @Xia-Weiwen
    Xia-Weiwen committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    00cfadd View commit details
    Browse the repository at this point in the history

  2. Update readme.md

    @Xia-Weiwen
    Xia-Weiwen committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    466d2f1 View commit details
    Browse the repository at this point in the history

  3. Debug CI failures (4)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    e970a4a View commit details
    Browse the repository at this point in the history

Commits on Oct 18, 2024

  1. Reimplement with nested tensor subclassing

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    5e2abbe View commit details
    Browse the repository at this point in the history

  2. Test torch.compile only with PyTorch >= 2.5

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    90d1b7d View commit details
    Browse the repository at this point in the history

  3. Debug CI failures (5)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    03d490b View commit details
    Browse the repository at this point in the history

  4. Debug CI failures (6)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    f595ed4 View commit details
    Browse the repository at this point in the history

  5. Debug CI failures (7)

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    2202f69 View commit details
    Browse the repository at this point in the history

  6. Use MovingAvg observer for activation; Update UT and readme

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    6ea8aa8 View commit details
    Browse the repository at this point in the history

  7. Revert changes to test_spinquant.py; refine readme

    @Xia-Weiwen
    Xia-Weiwen committed Oct 18, 2024
    Configuration menu
    Copy the full SHA
    fa1144c View commit details
    Browse the repository at this point in the history