Skip to content

Add ONNX exporter #26

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
Draft
santhnm2 wants to merge 63 commits into
base: main
Choose a base branch
from
Draft

Add ONNX exporter #26

santhnm2 wants to merge 63 commits into from

Conversation

@santhnm2
Copy link
Collaborator

@santhnm2 santhnm2 commented Jun 1, 2021

No description provided.

@siddharth-krishna
Projector: DistIR function -> per-rank function
8ac2c50
@siddharth-krishna
Fix bug: constructing variadic ops when specifying output_values
193dc84
@siddharth-krishna
Fix sequential executor docstring
c5bfbd9
@siddharth-krishna
A distributed PyTorch backend
d2f9e0a
@siddharth-krishna
Upgrade PyTorch version
d9cc787
@siddharth-krishna
Add test: one-weird-trick (matmul version)
a2a4cc2
@siddharth-krishna
Refactor run_multiprocess
7ab770d
@siddharth-krishna
Make run_multiprocess take Functions not nn.Modules
4e37ce3
@siddharth-krishna
End-of-file newlines
9576811
@siddharth-krishna
Black
c4ca1a0
@siddharth-krishna
Parametrize pytest test_owt
a095665
@siddharth-krishna
Revert "Fix bug: constructing variadic ops when specifying output_val…
3279ca1 
…ues"

This reverts commit 193dc84.
@siddharth-krishna
Fix Op constructor: better handling of variadic ops/pre-created outputs
c84c61d
@siddharth-krishna
Add support for Relu
bb604f6
@siddharth-krishna
Add DP test
8d8bbda
@siddharth-krishna
Backend: run and time on GPU
42751da
@siddharth-krishna
Refactor grid_search for interactive use
f73d33b
@siddharth-krishna
Timing code for CPUs
20bae75
@siddharth-krishna
Handle ops with multiple outputs
8e8513d
@siddharth-krishna
Add support for all MLP training ops
6f3ae59
@siddharth-krishna
Type inference: fix function name bug
49d385c
@siddharth-krishna
Prettyprint: support for printing FunctionMakers
2cce7c3
@siddharth-krishna
Fix backend op implementations
91f5ecf
@siddharth-krishna
Convert per-rank fns to Modules inside each thread
172a5db
@siddharth-krishna
Per-rank projector: remove types
2ed3cf7
@siddharth-krishna
Add some more tests
7eee9f9
@siddharth-krishna
Default number of repetitions = 1
78da37d
@siddharth-krishna
DHP transform: return separate init_fn and transformed fn
56e05dc
@siddharth-krishna
Run pytest with root dir included in PYTHONPATH
78bc1e9
@siddharth-krishna
Grid search: remove unused imports
8a288b6
siddharth-krishna and others added 27 commits May 5, 2021 17:07
@siddharth-krishna
Move new tensors to GPU in each op, outputs back to CPU
fd2e7b1
@siddharth-krishna
Add a mock multiprocess backend for debugging
31031cc
@siddharth-krishna
Use spawn start method for multiprocessing
5d99a62
@siddharth-krishna
Fix MLP DHP tests
2c8852a
@siddharth-krishna
Revert "Fix MLP DHP tests"
1d54fea 
This reverts commit 2c8852a.
@siddharth-krishna
Fix MLP DHP tests for real
18eaa08
@siddharth-krishna
Add code to plot grid search results
0ce2fdb
@siddharth-krishna
Don't use globals while multiprocessing
5588b87
@siddharth-krishna
Fix mock backend, use_gpu=False by default
7d714d4
@siddharth-krishna
Partial grid search on 4 devices
8646842
@siddharth-krishna
Support collectives between a subset of ranks
1db0736
@santhnm2
Bug fixes for distributed process groups (#24)
a79ab67 
Uses device IDs to index into `ctx.groups` for `send` and `recv` backend ops. Also ensures tuple of device IDs in `ctx.groups
@siddharth-krishna
Debugging MLP deadlock
0c0f7f5
@santhnm2
Fix collective projector
33cefce
@siddharth-krishna
Enable grid search test again
a5c8b34
@siddharth-krishna
Some comments, debugging code, cuda sync earlier
ca68ec4
@siddharth-krishna
Projector for gather
b31209d
@siddharth-krishna
Don't save input/outputs to file in torch backend
54ce0d2
@siddharth-krishna
Remove unnecessary global
5d1b63f
@siddharth-krishna
Free tensors after use
3466627
@siddharth-krishna
Map DistIR devices to pytorch backend ranks
7d16c90
@siddharth-krishna
Fix tests
8c4f5e1
@siddharth-krishna
Some documentation and cleanup
79d4a65
@siddharth-krishna
Fix comment
da7ff5d
@siddharth-krishna
Remove experiment code and dead code
72589b2
@santhnm2
Add ONNX exporter
849ab4e
@santhnm2
Remove singleton types
054ce0a
@santhnm2 santhnm2 changed the base branch from main to lowering June 1, 2021 04:59
Base automatically changed from lowering to main June 3, 2021 21:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@santhnm2 @siddharth-krishna