Skip to content

Add CSR (compressed sparse row) layout for sparse tensors #50937

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
Closed
v0dro wants to merge 272 commits into from
Closed

Add CSR (compressed sparse row) layout for sparse tensors #50937

v0dro wants to merge 272 commits into from

Conversation

@v0dro
Copy link
Contributor

@v0dro v0dro commented Jan 22, 2021

Implement compressed sparse row format. Derived from the GCS implementation at #44190

v0dro added 30 commits August 6, 2020 07:41
@v0dro
fixed code generation logic
d7c5f37
@v0dro
write manual constructor for sparse GCS
da25d86
@v0dro
basic construction of sparse GCS works
10afb3f
@v0dro
put dummy data inside member data
ee7b126
@v0dro
can construct a GCS tensor nicely without any safety and autograd
e126f93
@v0dro
merge master
06a23ee
@v0dro
update test
7477372
@v0dro
update GCS tests
96b698d
@v0dro
new tests for sparse GCS
66072f7
@v0dro
adding simple addition function to sparse GCS
370f612
@v0dro
update strides script
42a296e
@v0dro
Merge branch 'master' of github.com:pytorch/pytorch into sparse-csr-l…
72cf387 
…ayout
@v0dro
update sparse GCS tensor addition routines
1f1a1bf
@v0dro
adding dense conversion routine
1a35982
@v0dro
begin adding functions for conversion of GCS to dense co ordinates
b73dfc2
@v0dro
adding support for making strides
44897e3
@v0dro
update stride caluculation
13ad700
@v0dro
add more functions
1c57a0c
@v0dro
merge with master
73f4f23
@v0dro
update GCS to dense index conversion
99311e2
@v0dro
conversion to dense works
ed57303
@v0dro
add small addition method
17593dc
@v0dro
update master
d372f96
@v0dro
GCS matmul test
c67ddf9
@v0dro
Merge branch 'master' of github.com:pytorch/pytorch into sparse-csr-l…
c9fa58f 
…ayout
@v0dro
update small C++ function
1b111c5
@v0dro
update scaffold for spmv
fa3dc8d
@v0dro
new function for addmm functionality with and without out param
e5eafd6
@v0dro
update GCS tensor matvec
a2cfc6b
@v0dro
Merge branch 'master' of github.com:pytorch/pytorch into sparse-csr-l…
232d8ef 
…ayout
@peterbell10 peterbell10 mentioned this pull request Apr 21, 2021
Closed
This was referenced Apr 22, 2021
Closed
Open
@aocsa aocsa mentioned this pull request Apr 22, 2021
Closed
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update base for Update on "Modernize test-suite in sparse tensors"
e6f4fbb 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed here https://pearu.github.io/csr_tensor_support.html

#50937 (comment) - see modern COO testing
#50937 (comment) - don’t set default_dtype_type
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update on "Modernize test-suite in sparse tensors"
f1f15bd 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed here https://pearu.github.io/csr_tensor_support.html

#50937 (comment) - see modern COO testing
#50937 (comment) - don’t set default_dtype_type
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
@aocsa aocsa mentioned this pull request Apr 22, 2021
Closed
facebook-github-bot pushed a commit that referenced this pull request Apr 22, 2021
@peterbell10 @facebook-github-bot
Tell codegen that SparseCsrCUDA is cuda (#56602)
8b3bf98 
Summary:
Follow up to #50937, Fixes build failures in #56561

Currently SparseCsrCUDA is included in cpu build and also doesn't get code-generated device guards. This fixed both issues.

Pull Request resolved: #56602

Reviewed By: albanD

Differential Revision: D27921001

Pulled By: ezyang

fbshipit-source-id: 2b3b0b66d0a7c5ef96e0817d8852d511dd954ae4
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update base for Update on "Eliminate global usage of torch.set_defaul…
0be222f 
…t_dtype in sparse test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update on "Eliminate global usage of torch.set_default_dtype in spars…
d61aad9 
…e test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update base for Update on "Modernize test-suite in sparse tensors "
8f7bd02 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed here https://pearu.github.io/csr_tensor_support.html

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update base for Update on "Eliminate global usage of torch.set_defaul…
1cd63ae 
…t_dtype in sparse test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update on "Modernize test-suite in sparse tensors "
e2bb36d 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed here https://pearu.github.io/csr_tensor_support.html

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 22, 2021
@aocsa
Update on "Eliminate global usage of torch.set_default_dtype in spars…
2565070 
…e test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update base for Update on "Modernize test-suite in sparse tensors "
705fdd8 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update on "Modernize test-suite in sparse tensors "
8eaa2de 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update base for Update on "Eliminate global usage of torch.set_defaul…
d7a33c4 
…t_dtype in sparse test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update on "Eliminate global usage of torch.set_default_dtype in spars…
5074fa3 
…e test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update base for Update on "Modernize test-suite in sparse tensors "
adc3f61 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update base for Update on "Eliminate global usage of torch.set_defaul…
2264e1c 
…t_dtype in sparse test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update on "Eliminate global usage of torch.set_default_dtype in spars…
9f1c867 
…e test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 23, 2021
@aocsa
Update on "Modernize test-suite in sparse tensors "
2660dc2 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 26, 2021
@aocsa
Update base for Update on "Modernize test-suite in sparse tensors "
ed49740 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 26, 2021
@aocsa
Update base for Update on "Eliminate global usage of torch.set_defaul…
e1cc221 
…t_dtype in sparse test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 26, 2021
@aocsa
Update on "Eliminate global usage of torch.set_default_dtype in spars…
5dcdc01 
…e test"


Fixes for  gh-56369

Note:
The following request  about testing 

#50937 (comment) - don’t set default_dtype_type

is being solved in this PR



Differential Revision: [D27913266](https://our.internmc.facebook.com/intern/diff/D27913266)

[ghstack-poisoned]
aocsa added a commit that referenced this pull request Apr 26, 2021
@aocsa
Update on "Modernize test-suite in sparse tensors "
2cd9c73 
Fixes for gh-56371 and gh-56369

Differential Revision: [D27913212](https://our.internmc.facebook.com/intern/diff/D27913212)

Note: 
The following requests about testing listed [here](https://pearu.github.io/csr_tensor_support.html) 

#50937 (comment) - see modern COO testing
#50937 (comment) - avoid using numpy
#50937 (comment) - inefficient CSR samples

are being solved in this PR

[ghstack-poisoned]
@pearu pearu mentioned this pull request Apr 30, 2021
Closed
krshrimali pushed a commit to krshrimali/pytorch that referenced this pull request May 19, 2021
@v0dro
Add CSR (compressed sparse row) layout for sparse tensors (pytorch#50937
a974a2b 
)

Summary:
Implement compressed sparse row format. Derived from the GCS implementation at pytorch#44190

Pull Request resolved: pytorch#50937

Reviewed By: mrshenli

Differential Revision: D27439865

Pulled By: ezyang

fbshipit-source-id: 3ba3dcb9679505b980ff6a5f513e913bbae2fb1d
krshrimali pushed a commit to krshrimali/pytorch that referenced this pull request May 19, 2021
@peterbell10
Tell codegen that SparseCsrCUDA is cuda (pytorch#56602)
5822a37 
Summary:
Follow up to pytorch#50937, Fixes build failures in pytorch#56561

Currently SparseCsrCUDA is included in cpu build and also doesn't get code-generated device guards. This fixed both issues.

Pull Request resolved: pytorch#56602

Reviewed By: albanD

Differential Revision: D27921001

Pulled By: ezyang

fbshipit-source-id: 2b3b0b66d0a7c5ef96e0817d8852d511dd954ae4
@malfet malfet mentioned this pull request Mar 22, 2023
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ezyang ezyang ezyang approved these changes

@rgommers rgommers rgommers left review comments

@pearu pearu pearu approved these changes

@albanD albanD Awaiting requested review from albanD

@soulitzer soulitzer Awaiting requested review from soulitzer

@BowenBao BowenBao Awaiting requested review from BowenBao

@fmassa fmassa Awaiting requested review from fmassa

@glaringlee glaringlee Awaiting requested review from glaringlee

@H-Huang H-Huang Awaiting requested review from H-Huang

@jbschlosser jbschlosser Awaiting requested review from jbschlosser

@mingzhe09088 mingzhe09088 Awaiting requested review from mingzhe09088

@mrshenli mrshenli Awaiting requested review from mrshenli

@neginraoof neginraoof Awaiting requested review from neginraoof

@pritamdamania87 pritamdamania87 Awaiting requested review from pritamdamania87

@rohan-varma rohan-varma Awaiting requested review from rohan-varma

@wayi1 wayi1 Awaiting requested review from wayi1

@soumith soumith Awaiting requested review from soumith

@spandantiwari spandantiwari Awaiting requested review from spandantiwari

@zhaojuanmao zhaojuanmao Awaiting requested review from zhaojuanmao

@wenleix wenleix Awaiting requested review from wenleix

+1 more reviewer

@aocsa aocsa aocsa approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

cla signed Merged module: sparse Related to torch.sparse open source triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.