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

Poisson NLL loss #1774

Closed
vlasenkov opened this issue Jun 11, 2017 · 2 comments · Fixed by #1779
Closed

Poisson NLL loss #1774

vlasenkov opened this issue Jun 11, 2017 · 2 comments · Fixed by #1779

Comments

@vlasenkov
Copy link
Contributor

vlasenkov commented Jun 11, 2017 •
edited

There is no log-likelihood loss with underlying target's Poisson distribution. Would be great to add. I can do it.
@vlasenkov vlasenkov changed the title Poisson loss function Poisson NLL loss Jun 11, 2017
@vlasenkov
Copy link
Contributor Author

I'm not sure about where implementations of different losses are stored. To create new loss I need:

  1. to add new class to torch/nn/modules/loss.py
  2. then register implementation of the loss somewhere in torch/nn/_functions/thnn
  3. But what are the locations for these implementations? torch/legacy or torch/nn/functional.py or torch/nn/_functions/loss.py or some C code?

@fmassa
Copy link
Member

fmassa commented Jun 11, 2017

The C/Cuda implementation can be found in libs/THNN and libs/THCUNN, which are automatically wrapped in python via torch/nn/_functions/thnn/auto.py. If the function can be trivially implemented using autograd, the implementation could go to torch/nn/functional, followed by a wrapper in the modules. If you want to write the backprop by hand and in python (for example for optimization reasons), then you could write it in torch/nn/_functions/loss.py for example.

@vlasenkov vlasenkov mentioned this issue Jun 12, 2017
Merged
@nailimixaM nailimixaM mentioned this issue Nov 27, 2020
Closed
IvanYashchuk pushed a commit to IvanYashchuk/pytorch that referenced this issue Jun 27, 2022
@naoyam
Issue 1770 (pytorch#1774)
1b65299 
* Fix issue pytorch#1770

* Adds assertions to avoid segfault
pytorchmergebot pushed a commit that referenced this issue Jul 13, 2022
@jjsjann123 @pytorchmergebot
[WIP] Upstream push 0627 (#80355)
8d753c8 
Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/

Code changes includes:

- TransformPropagator refactor: switched to Dijkstra instead of exhaustive enumeration on all possible paths to reduce compilation time on transform propagation;
- Indexing refactor: remove reference tensor creation in all tensor indexing logic (#1690)
- (more) generic grouped grid reduction kernel;
- Minor parser/fuser patches:
  1. zero-dim tensor reduction support
  3. no-op binary removal within fused graph
  4. expand supported in fusion

Squashed commits to WAR github API
Commits that's actually in this PR from the devel branch:

```
a054b3e Refactor TransormPropagator to allow specifying a position and propagating to part of the DAG (#1775)
d67e1cd Indexing refactor stage 1: remove reference tensor creation in all tensor indexing logic (#1690)
1b65299 Issue 1770 (#1774)
35b0427 Avoid compilation errors like below: (#1773)
452c773 Ignore reductions of zero-dim tensors per PyTorch conventions (#1771)
31d6c56 TransformPropagator refactor (#1769)
570c5a8 Merge pull request #1767 from csarofeen/upstream_merge_0621
9d6c3d8 merging upstream 61305cd
0ed815f New TransformPropagator algorithm (#1763)
6c19520 no-op binary removal (#1764)
ec7fa41 Proper propagation of IterType (#1762)
b263562 Fix dimensionality check (#1759)
2d6343f More generic grouped grid reduction kernel (#1740)
64e2b56 [nvfuser] prevent spamming warning message (#77777) (#1758)
0c43162 [nvFuser] Improving bitwise ops support (#77158) (#1757)
b93a147 Parser expand (#1754)
```

RUN_TORCHBENCH: nvfuser
Pull Request resolved: #80355
Approved by: https://github.com/davidberard98
facebook-github-bot pushed a commit that referenced this issue Jul 13, 2022
@jjsjann123 @facebook-github-bot
[WIP] Upstream push 0627 (#80355)
c712397 
Summary:
Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/

Code changes includes:

- TransformPropagator refactor: switched to Dijkstra instead of exhaustive enumeration on all possible paths to reduce compilation time on transform propagation;
- Indexing refactor: remove reference tensor creation in all tensor indexing logic (#1690)
- (more) generic grouped grid reduction kernel;
- Minor parser/fuser patches:
  1. zero-dim tensor reduction support
  3. no-op binary removal within fused graph
  4. expand supported in fusion

Squashed commits to WAR github API
Commits that's actually in this PR from the devel branch:

```
a054b3e Refactor TransormPropagator to allow specifying a position and propagating to part of the DAG (#1775)
d67e1cd Indexing refactor stage 1: remove reference tensor creation in all tensor indexing logic (#1690)
1b65299 Issue 1770 (#1774)
35b0427 Avoid compilation errors like below: (#1773)
452c773 Ignore reductions of zero-dim tensors per PyTorch conventions (#1771)
31d6c56 TransformPropagator refactor (#1769)
570c5a8 Merge pull request #1767 from csarofeen/upstream_merge_0621
9d6c3d8 merging upstream 61305cd
0ed815f New TransformPropagator algorithm (#1763)
6c19520 no-op binary removal (#1764)
ec7fa41 Proper propagation of IterType (#1762)
b263562 Fix dimensionality check (#1759)
2d6343f More generic grouped grid reduction kernel (#1740)
64e2b56 [nvfuser] prevent spamming warning message (#77777) (#1758)
0c43162 [nvFuser] Improving bitwise ops support (#77158) (#1757)
b93a147 Parser expand (#1754)
```

RUN_TORCHBENCH: nvfuser

Pull Request resolved: #80355

Reviewed By: qihqi

Differential Revision: D37573400

Pulled By: davidberard98

fbshipit-source-id: 52ab68d89ec01ef61f69f5abeb18c9d3a312aa64
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

adds poisson NLL loss vlasenkov/pytorch
2 participants
@fmassa @vlasenkov