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

Some follow up fixes for quant primitives #220

Merged
merged 3 commits into from
May 8, 2024
Merged

Some follow up fixes for quant primitives #220

merged 3 commits into from
May 8, 2024

Conversation

jerryzh168
Copy link
Contributor

Summary:
att

Test Plan:
python test/quantization/test_quant_primitives.py -k test_raises

Reviewers:

Subscribers:

Tasks:

Tags:

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label May 6, 2024
@jerryzh168
Some follow up fixes for quant primitives
949d816 
Summary:
att

Test Plan:
python test/quantization/test_quant_primitives.py -k test_raises

Reviewers:

Subscribers:

Tasks:

Tags:
@pytorch-bot PyTorch Bot
Copy link

pytorch-bot bot commented May 7, 2024 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/220

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 2e69d6c with merge base 9849360 (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@jerryzh168 jerryzh168 merged commit 63c5ac5 into pytorch:main May 8, 2024
13 checks passed
HDCharles added a commit that referenced this pull request May 8, 2024
@HDCharles @jeromeku
@jerryzh168 @msaroufim @svekars
Composing autoquant with compile (#175)
f6d56ca 
* Composing autoquant with compile

Summary:

this PR rewrites how torchao.autoquant works so that it works with
torch.compile. Previously you had to do:

torchao.autoquant(model, input)
mod=torch.compile(model)
mod(input)

now you can do
torchao.autoquant(torch.compile(model))
model(input)

The new method works with/without compile. Also this is BC so the old
path also works.

We use a forward_prehook to intercept the model call before
torch.compile tracing occurs at which point we do the autoquantization
and clean up all remaining hooks before passing things off to the
normal torch.compile tracing functionality.

note: in the case of multiple inputs, you can also do:

model.forward_log_only(input) to run the model forward with autoquant
shape logging and prevent the torch.compile tracing/autoquant
quantization from occuring.

Test Plan: python test/integration/test_integration.py -k "autoquant"

Reviewers:

Subscribers:

Tasks:

Tags:

* Fused DoRA kernels (#216)

* add dora kernels

* allowing error_on_unseen in autoquant func

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* Unified AffineQuantizedTensor subclass (#214)

Summary:
Creatd a `AffineQuantizedTensor` subclass that works for both weight and input (for dynamic quantization), for all granularities (levering the recently added choose_qparams_affine, quantize_affine
and dequantize_affine ops)

only verified for 8da4w right now, we can make it work for other types of quantization (mostly the operator dispatching part) later

Test Plan:
python test/quantization/test_quant_api.py -k test_quantized_tensor_subclass_8da4w

Reviewers:

Subscribers:

Tasks:

Tags:

Co-authored-by: Mark Saroufim <marksaroufim@meta.com>

* add expecttest to requirements.txt (#225)

* add expecttest to requirements.txt

* update

* Install dev-requirements.txt in doc build (#224)

Install dev-requirements.txt

---------

Co-authored-by: Mark Saroufim <marksaroufim@meta.com>

* Fix an error in subclass impl (#226)

Summary:
Accidently changed the device check code for old subclass instead of the new one, forgot to fix before landing

Test Plan:
CI

Reviewers:

Subscribers:

Tasks:

Tags:

* update readme.md

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* trying to fix the error in CI on cleanup hooks

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* correct docs

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* Some follow up fixes for quant primitives (#220)

Summary:
att

Test Plan:
python test/quantization/test_quant_primitives.py -k test_raises

Reviewers:

Subscribers:

Tasks:

Tags:

* Composing autoquant with compile

Summary:

this PR rewrites how torchao.autoquant works so that it works with
torch.compile. Previously you had to do:

torchao.autoquant(model, input)
mod=torch.compile(model)
mod(input)

now you can do
torchao.autoquant(torch.compile(model))
model(input)

The new method works with/without compile. Also this is BC so the old
path also works.

We use a forward_prehook to intercept the model call before
torch.compile tracing occurs at which point we do the autoquantization
and clean up all remaining hooks before passing things off to the
normal torch.compile tracing functionality.

note: in the case of multiple inputs, you can also do:

model.forward_log_only(input) to run the model forward with autoquant
shape logging and prevent the torch.compile tracing/autoquant
quantization from occuring.

Test Plan: python test/integration/test_integration.py -k "autoquant"

Reviewers:

Subscribers:

Tasks:

Tags:

* allowing error_on_unseen in autoquant func

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* update readme.md

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* trying to fix the error in CI on cleanup hooks

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* correct docs

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

---------

Co-authored-by: jeromeku <jerome.ku@gmail.com>
Co-authored-by: Jerry Zhang <jerryzh168@gmail.com>
Co-authored-by: Mark Saroufim <marksaroufim@meta.com>
Co-authored-by: Svetlana Karslioglu <svekars@meta.com>
dbyoung18 pushed a commit to dbyoung18/ao that referenced this pull request Jul 31, 2024
@jerryzh168
Some follow up fixes for quant primitives (pytorch#220)
2e252d0 
Summary:
att

Test Plan:
python test/quantization/test_quant_primitives.py -k test_raises

Reviewers:

Subscribers:

Tasks:

Tags:
dbyoung18 pushed a commit to dbyoung18/ao that referenced this pull request Jul 31, 2024
@HDCharles @jeromeku
@jerryzh168 @msaroufim @svekars
Composing autoquant with compile (pytorch#175)
32b6f0a 
* Composing autoquant with compile

Summary:

this PR rewrites how torchao.autoquant works so that it works with
torch.compile. Previously you had to do:

torchao.autoquant(model, input)
mod=torch.compile(model)
mod(input)

now you can do
torchao.autoquant(torch.compile(model))
model(input)

The new method works with/without compile. Also this is BC so the old
path also works.

We use a forward_prehook to intercept the model call before
torch.compile tracing occurs at which point we do the autoquantization
and clean up all remaining hooks before passing things off to the
normal torch.compile tracing functionality.

note: in the case of multiple inputs, you can also do:

model.forward_log_only(input) to run the model forward with autoquant
shape logging and prevent the torch.compile tracing/autoquant
quantization from occuring.

Test Plan: python test/integration/test_integration.py -k "autoquant"

Reviewers:

Subscribers:

Tasks:

Tags:

* Fused DoRA kernels (pytorch#216)

* add dora kernels

* allowing error_on_unseen in autoquant func

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* Unified AffineQuantizedTensor subclass (pytorch#214)

Summary:
Creatd a `AffineQuantizedTensor` subclass that works for both weight and input (for dynamic quantization), for all granularities (levering the recently added choose_qparams_affine, quantize_affine
and dequantize_affine ops)

only verified for 8da4w right now, we can make it work for other types of quantization (mostly the operator dispatching part) later

Test Plan:
python test/quantization/test_quant_api.py -k test_quantized_tensor_subclass_8da4w

Reviewers:

Subscribers:

Tasks:

Tags:

Co-authored-by: Mark Saroufim <marksaroufim@meta.com>

* add expecttest to requirements.txt (pytorch#225)

* add expecttest to requirements.txt

* update

* Install dev-requirements.txt in doc build (pytorch#224)

Install dev-requirements.txt

---------

Co-authored-by: Mark Saroufim <marksaroufim@meta.com>

* Fix an error in subclass impl (pytorch#226)

Summary:
Accidently changed the device check code for old subclass instead of the new one, forgot to fix before landing

Test Plan:
CI

Reviewers:

Subscribers:

Tasks:

Tags:

* update readme.md

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* trying to fix the error in CI on cleanup hooks

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* correct docs

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* Some follow up fixes for quant primitives (pytorch#220)

Summary:
att

Test Plan:
python test/quantization/test_quant_primitives.py -k test_raises

Reviewers:

Subscribers:

Tasks:

Tags:

* Composing autoquant with compile

Summary:

this PR rewrites how torchao.autoquant works so that it works with
torch.compile. Previously you had to do:

torchao.autoquant(model, input)
mod=torch.compile(model)
mod(input)

now you can do
torchao.autoquant(torch.compile(model))
model(input)

The new method works with/without compile. Also this is BC so the old
path also works.

We use a forward_prehook to intercept the model call before
torch.compile tracing occurs at which point we do the autoquantization
and clean up all remaining hooks before passing things off to the
normal torch.compile tracing functionality.

note: in the case of multiple inputs, you can also do:

model.forward_log_only(input) to run the model forward with autoquant
shape logging and prevent the torch.compile tracing/autoquant
quantization from occuring.

Test Plan: python test/integration/test_integration.py -k "autoquant"

Reviewers:

Subscribers:

Tasks:

Tags:

* allowing error_on_unseen in autoquant func

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* update readme.md

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* trying to fix the error in CI on cleanup hooks

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

* correct docs

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

---------

Co-authored-by: jeromeku <jerome.ku@gmail.com>
Co-authored-by: Jerry Zhang <jerryzh168@gmail.com>
Co-authored-by: Mark Saroufim <marksaroufim@meta.com>
Co-authored-by: Svetlana Karslioglu <svekars@meta.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@kimishpatel kimishpatel kimishpatel approved these changes

@msaroufim msaroufim Awaiting requested review from msaroufim

@HDCharles HDCharles Awaiting requested review from HDCharles

Assignees
No one assigned
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@jerryzh168 @kimishpatel @facebook-github-bot